Skip to contentSkip to conversation

Add generated latest Intel events and metrics

PATCH Superseded Ian Rogers 4 participants 42 messages Jul 22, 2022 – Jul 27, 2022 1 message not in the archive Page 1 of 2

Subsystems (2)

v1 of 3 Superseded Latest is v3 →

lore
Share & tools
lore

Ian Rogers

author reviewer
#0 lore raw

cover Add generated latest Intel events and metrics

The goal of this patch series is to align the json events for Intel platforms with those generated by:
https://github.com/intel/event-converter-for-linux-perf/blob/master/download_and_gen.py
This script takes the latest event json and TMA metrics from:
https://download.01.org/perfmon/ and adds to these metrics, in particular uncore ones, from: https://github.com/intel/perfmon-metrics
The cpu_operating_frequency metric assumes the presence of the system_tsc_freq literal posted/reviewed in:
https://lore.kernel.org/lkml/[email protected]/

Some fixes were needed to the script for generating the json and are contained in this pull request:
https://github.com/intel/event-converter-for-linux-perf/pull/15

The json files were first downloaded before being used to generate the perf json files. This fixes non-ascii characters for (R) and (TM) in the source json files. This can be reproduced with:

$ download_and_gen.py --hermetic-download --outdir data
$ download_and_gen.py --url=file://`pwd`/data/01 --metrics-url=file://`pwd`/data/github

A minor correction is made in the generated json of:
tools/perf/pmu-events/arch/x86/ivytown/uncore-other.json changing "\\Inbound\\" to just "Inbound" to avoid compilation errors caused by \I.

The elkhartlake metrics file is basic and not generated by scripts. It is retained here although it causes a difference from the generated files.

The mapfile.csv is the third and final difference from the generated version due to a bug in 01.org's models for icelake. The existing models are preferred and retained.

As well as the #system_tsc_freq being necessary, a test change is added here fixing an issue with fake PMU testing exposed in the new/updated metrics.

Compared to the previous json, additional changes are the inclusion of basic meteorlake events and the renaming of tremontx to snowridgex. The new metrics contribute to the size, but a large contribution is the inclusion of previously ungenerated and experimental uncore events.

Ian Rogers (31):
perf test: Avoid sysfs state affecting fake events perf vendor events: Update Intel broadwellx perf vendor events: Update Intel broadwell perf vendor events: Update Intel broadwellde perf vendor events: Update Intel alderlake perf vendor events: Update bonnell mapfile.csv perf vendor events: Update Intel cascadelakex perf vendor events: Update Intel elkhartlake perf vendor events: Update goldmont mapfile.csv perf vendor events: Update goldmontplus mapfile.csv perf vendor events: Update Intel haswell perf vendor events: Update Intel haswellx perf vendor events: Update Intel icelake perf vendor events: Update Intel icelakex perf vendor events: Update Intel ivybridge perf vendor events: Update Intel ivytown perf vendor events: Update Intel jaketown perf vendor events: Update Intel knightslanding perf vendor events: Add Intel meteorlake perf vendor events: Update Intel nehalemep perf vendor events: Update Intel nehalemex perf vendor events: Update Intel sandybridge perf vendor events: Update Intel sapphirerapids perf vendor events: Update Intel silvermont perf vendor events: Update Intel skylake perf vendor events: Update Intel skylakex perf vendor events: Update Intel snowridgex perf vendor events: Update Intel tigerlake perf vendor events: Update Intel westmereep-dp perf vendor events: Update Intel westmereep-sp perf vendor events: Update Intel westmereex

 .../arch/x86/alderlake/adl-metrics.json       |     4 +-
 .../pmu-events/arch/x86/alderlake/cache.json  |   178 +-
 .../arch/x86/alderlake/floating-point.json    |    19 +-
 .../arch/x86/alderlake/frontend.json          |    38 +-
 .../pmu-events/arch/x86/alderlake/memory.json |    40 +-
 .../pmu-events/arch/x86/alderlake/other.json  |    97 +-
 .../arch/x86/alderlake/pipeline.json          |   507 +-
 .../arch/x86/alderlake/uncore-other.json      |     2 +-
 .../arch/x86/alderlake/virtual-memory.json    |    63 +-
 .../pmu-events/arch/x86/bonnell/cache.json    |     2 +-
 .../arch/x86/bonnell/floating-point.json      |     2 +-
 .../pmu-events/arch/x86/bonnell/frontend.json |     2 +-
 .../pmu-events/arch/x86/bonnell/memory.json   |     2 +-
 .../pmu-events/arch/x86/bonnell/other.json    |     2 +-
 .../pmu-events/arch/x86/bonnell/pipeline.json |     2 +-
 .../arch/x86/bonnell/virtual-memory.json      |     2 +-
 .../arch/x86/broadwell/bdw-metrics.json       |   130 +-
 .../pmu-events/arch/x86/broadwell/cache.json  |     2 +-
 .../arch/x86/broadwell/floating-point.json    |     2 +-
 .../arch/x86/broadwell/frontend.json          |     2 +-
 .../pmu-events/arch/x86/broadwell/memory.json |     2 +-
 .../pmu-events/arch/x86/broadwell/other.json  |     2 +-
 .../arch/x86/broadwell/pipeline.json          |     2 +-
 .../arch/x86/broadwell/uncore-cache.json      |   152 +
 .../arch/x86/broadwell/uncore-other.json      |    82 +
 .../pmu-events/arch/x86/broadwell/uncore.json |   278 -
 .../arch/x86/broadwell/virtual-memory.json    |     2 +-
 .../arch/x86/broadwellde/bdwde-metrics.json   |   136 +-
 .../arch/x86/broadwellde/cache.json           |     2 +-
 .../arch/x86/broadwellde/floating-point.json  |     2 +-
 .../arch/x86/broadwellde/frontend.json        |     2 +-
 .../arch/x86/broadwellde/memory.json          |     2 +-
 .../arch/x86/broadwellde/other.json           |     2 +-
 .../arch/x86/broadwellde/pipeline.json        |     2 +-
 .../arch/x86/broadwellde/uncore-cache.json    |  3818 ++-
 .../arch/x86/broadwellde/uncore-memory.json   |  2867 +-
 .../arch/x86/broadwellde/uncore-other.json    |  1246 +
 .../arch/x86/broadwellde/uncore-power.json    |   492 +-
 .../arch/x86/broadwellde/virtual-memory.json  |     2 +-
 .../arch/x86/broadwellx/bdx-metrics.json      |   570 +-
 .../pmu-events/arch/x86/broadwellx/cache.json |    22 +-
 .../arch/x86/broadwellx/floating-point.json   |     9 +-
 .../arch/x86/broadwellx/frontend.json         |     2 +-
 .../arch/x86/broadwellx/memory.json           |    39 +-
 .../pmu-events/arch/x86/broadwellx/other.json |     2 +-
 .../arch/x86/broadwellx/pipeline.json         |     4 +-
 .../arch/x86/broadwellx/uncore-cache.json     |  3788 ++-
 .../x86/broadwellx/uncore-interconnect.json   |  1438 +-
 .../arch/x86/broadwellx/uncore-memory.json    |  2849 +-
 .../arch/x86/broadwellx/uncore-other.json     |  3252 ++
 .../arch/x86/broadwellx/uncore-power.json     |   437 +-
 .../arch/x86/broadwellx/virtual-memory.json   |     2 +-
 .../arch/x86/cascadelakex/cache.json          |     8 +-
 .../arch/x86/cascadelakex/clx-metrics.json    |   724 +-
 .../arch/x86/cascadelakex/floating-point.json |     2 +-
 .../arch/x86/cascadelakex/frontend.json       |     2 +-
 .../arch/x86/cascadelakex/other.json          |    63 +
 .../arch/x86/cascadelakex/pipeline.json       |    11 +
 .../arch/x86/cascadelakex/uncore-memory.json  |     9 +
 .../arch/x86/cascadelakex/uncore-other.json   |   697 +-
 .../arch/x86/cascadelakex/virtual-memory.json |     2 +-
 .../arch/x86/elkhartlake/cache.json           |   956 +-
 .../arch/x86/elkhartlake/floating-point.json  |    19 +-
 .../arch/x86/elkhartlake/frontend.json        |    34 +-
 .../arch/x86/elkhartlake/memory.json          |   388 +-
 .../arch/x86/elkhartlake/other.json           |   527 +-
 .../arch/x86/elkhartlake/pipeline.json        |   203 +-
 .../arch/x86/elkhartlake/virtual-memory.json  |   151 +-
 .../pmu-events/arch/x86/goldmont/cache.json   |     2 +-
 .../arch/x86/goldmont/floating-point.json     |     2 +-
 .../arch/x86/goldmont/frontend.json           |     2 +-
 .../pmu-events/arch/x86/goldmont/memory.json  |     2 +-
 .../arch/x86/goldmont/pipeline.json           |     2 +-
 .../arch/x86/goldmont/virtual-memory.json     |     2 +-
 .../arch/x86/goldmontplus/cache.json          |     2 +-
 .../arch/x86/goldmontplus/floating-point.json |     2 +-
 .../arch/x86/goldmontplus/frontend.json       |     2 +-
 .../arch/x86/goldmontplus/memory.json         |     2 +-
 .../arch/x86/goldmontplus/pipeline.json       |     2 +-
 .../arch/x86/goldmontplus/virtual-memory.json |     2 +-
 .../pmu-events/arch/x86/haswell/cache.json    |    78 +-
 .../arch/x86/haswell/floating-point.json      |     2 +-
 .../pmu-events/arch/x86/haswell/frontend.json |     2 +-
 .../arch/x86/haswell/hsw-metrics.json         |    85 +-
 .../pmu-events/arch/x86/haswell/memory.json   |    75 +-
 .../pmu-events/arch/x86/haswell/other.json    |     2 +-
 .../pmu-events/arch/x86/haswell/pipeline.json |     9 +-
 .../arch/x86/haswell/uncore-other.json        |     7 +-
 .../arch/x86/haswell/virtual-memory.json      |     2 +-
 .../pmu-events/arch/x86/haswellx/cache.json   |    44 +-
 .../arch/x86/haswellx/floating-point.json     |     2 +-
 .../arch/x86/haswellx/frontend.json           |     2 +-
 .../arch/x86/haswellx/hsx-metrics.json        |    85 +-
 .../pmu-events/arch/x86/haswellx/memory.json  |    52 +-
 .../pmu-events/arch/x86/haswellx/other.json   |     2 +-
 .../arch/x86/haswellx/pipeline.json           |     9 +-
 .../arch/x86/haswellx/uncore-cache.json       |  3779 ++-
 .../x86/haswellx/uncore-interconnect.json     |  1430 +-
 .../arch/x86/haswellx/uncore-memory.json      |  2839 +-
 .../arch/x86/haswellx/uncore-other.json       |  3170 ++
 .../arch/x86/haswellx/uncore-power.json       |   477 +-
 .../arch/x86/haswellx/virtual-memory.json     |     2 +-
 .../pmu-events/arch/x86/icelake/cache.json    |     8 +-
 .../arch/x86/icelake/floating-point.json      |     2 +-
 .../pmu-events/arch/x86/icelake/frontend.json |     2 +-
 .../arch/x86/icelake/icl-metrics.json         |   126 +-
 .../arch/x86/icelake/uncore-other.json        |    31 +
 .../arch/x86/icelake/virtual-memory.json      |     2 +-
 .../pmu-events/arch/x86/icelakex/cache.json   |    28 +-
 .../arch/x86/icelakex/floating-point.json     |     2 +-
 .../arch/x86/icelakex/frontend.json           |     2 +-
 .../arch/x86/icelakex/icx-metrics.json        |   691 +-
 .../pmu-events/arch/x86/icelakex/memory.json  |     6 +-
 .../pmu-events/arch/x86/icelakex/other.json   |    51 +-
 .../arch/x86/icelakex/pipeline.json           |    12 +
 .../arch/x86/icelakex/virtual-memory.json     |     2 +-
 .../pmu-events/arch/x86/ivybridge/cache.json  |     2 +-
 .../arch/x86/ivybridge/floating-point.json    |     2 +-
 .../arch/x86/ivybridge/frontend.json          |     2 +-
 .../arch/x86/ivybridge/ivb-metrics.json       |    94 +-
 .../pmu-events/arch/x86/ivybridge/memory.json |     2 +-
 .../pmu-events/arch/x86/ivybridge/other.json  |     2 +-
 .../arch/x86/ivybridge/pipeline.json          |     4 +-
 .../arch/x86/ivybridge/uncore-other.json      |     2 +-
 .../arch/x86/ivybridge/virtual-memory.json    |     2 +-
 .../pmu-events/arch/x86/ivytown/cache.json    |     2 +-
 .../arch/x86/ivytown/floating-point.json      |     2 +-
 .../pmu-events/arch/x86/ivytown/frontend.json |     2 +-
 .../arch/x86/ivytown/ivt-metrics.json         |    94 +-
 .../pmu-events/arch/x86/ivytown/memory.json   |     2 +-
 .../pmu-events/arch/x86/ivytown/other.json    |     2 +-
 .../arch/x86/ivytown/uncore-cache.json        |  3495 ++-
 .../arch/x86/ivytown/uncore-interconnect.json |  1750 +-
 .../arch/x86/ivytown/uncore-memory.json       |  1775 +-
 .../arch/x86/ivytown/uncore-other.json        |  2411 ++
 .../arch/x86/ivytown/uncore-power.json        |   696 +-
 .../arch/x86/ivytown/virtual-memory.json      |     2 +-
 .../pmu-events/arch/x86/jaketown/cache.json   |     2 +-
 .../arch/x86/jaketown/floating-point.json     |     2 +-
 .../arch/x86/jaketown/frontend.json           |     2 +-
 .../arch/x86/jaketown/jkt-metrics.json        |    11 +-
 .../pmu-events/arch/x86/jaketown/memory.json  |     2 +-
 .../pmu-events/arch/x86/jaketown/other.json   |     2 +-
 .../arch/x86/jaketown/pipeline.json           |    16 +-
 .../arch/x86/jaketown/uncore-cache.json       |  1960 +-
 .../x86/jaketown/uncore-interconnect.json     |   824 +-
 .../arch/x86/jaketown/uncore-memory.json      |   445 +-
 .../arch/x86/jaketown/uncore-other.json       |  1551 +
 .../arch/x86/jaketown/uncore-power.json       |   362 +-
 .../arch/x86/jaketown/virtual-memory.json     |     2 +-
 .../arch/x86/knightslanding/cache.json        |     2 +-
 .../x86/knightslanding/floating-point.json    |     2 +-
 .../arch/x86/knightslanding/frontend.json     |     2 +-
 .../arch/x86/knightslanding/memory.json       |     2 +-
 .../arch/x86/knightslanding/pipeline.json     |     2 +-
 .../x86/knightslanding/uncore-memory.json     |    42 -
 .../arch/x86/knightslanding/uncore-other.json |  3890 +++
 .../x86/knightslanding/virtual-memory.json    |     2 +-
 tools/perf/pmu-events/arch/x86/mapfile.csv    |    74 +-
 .../pmu-events/arch/x86/meteorlake/cache.json |   262 +
 .../arch/x86/meteorlake/frontend.json         |    24 +
 .../arch/x86/meteorlake/memory.json           |   185 +
 .../pmu-events/arch/x86/meteorlake/other.json |    46 +
 .../arch/x86/meteorlake/pipeline.json         |   254 +
 .../arch/x86/meteorlake/virtual-memory.json   |    46 +
 .../pmu-events/arch/x86/nehalemep/cache.json  |    14 +-
 .../arch/x86/nehalemep/floating-point.json    |     2 +-
 .../arch/x86/nehalemep/frontend.json          |     2 +-
 .../pmu-events/arch/x86/nehalemep/memory.json |     6 +-
 .../arch/x86/nehalemep/virtual-memory.json    |     2 +-
 .../pmu-events/arch/x86/nehalemex/cache.json  |  2974 +-
 .../arch/x86/nehalemex/floating-point.json    |   182 +-
 .../arch/x86/nehalemex/frontend.json          |    20 +-
 .../pmu-events/arch/x86/nehalemex/memory.json |   672 +-
 .../pmu-events/arch/x86/nehalemex/other.json  |   170 +-
 .../arch/x86/nehalemex/pipeline.json          |   830 +-
 .../arch/x86/nehalemex/virtual-memory.json    |    92 +-
 .../arch/x86/sandybridge/cache.json           |     2 +-
 .../arch/x86/sandybridge/floating-point.json  |     2 +-
 .../arch/x86/sandybridge/frontend.json        |     4 +-
 .../arch/x86/sandybridge/memory.json          |     2 +-
 .../arch/x86/sandybridge/other.json           |     2 +-
 .../arch/x86/sandybridge/pipeline.json        |    10 +-
 .../arch/x86/sandybridge/snb-metrics.json     |    11 +-
 .../arch/x86/sandybridge/uncore-other.json    |     2 +-
 .../arch/x86/sandybridge/virtual-memory.json  |     2 +-
 .../arch/x86/sapphirerapids/cache.json        |   135 +-
 .../x86/sapphirerapids/floating-point.json    |     6 +
 .../arch/x86/sapphirerapids/frontend.json     |    16 +
 .../arch/x86/sapphirerapids/memory.json       |    23 +-
 .../arch/x86/sapphirerapids/other.json        |    68 +-
 .../arch/x86/sapphirerapids/pipeline.json     |    99 +-
 .../arch/x86/sapphirerapids/spr-metrics.json  |   566 +-
 .../arch/x86/sapphirerapids/uncore-other.json |     9 -
 .../x86/sapphirerapids/virtual-memory.json    |    20 +
 .../pmu-events/arch/x86/silvermont/cache.json |     2 +-
 .../arch/x86/silvermont/floating-point.json   |     2 +-
 .../arch/x86/silvermont/frontend.json         |     2 +-
 .../arch/x86/silvermont/memory.json           |     2 +-
 .../pmu-events/arch/x86/silvermont/other.json |     2 +-
 .../arch/x86/silvermont/pipeline.json         |     2 +-
 .../arch/x86/silvermont/virtual-memory.json   |     2 +-
 .../arch/x86/skylake/floating-point.json      |     2 +-
 .../pmu-events/arch/x86/skylake/frontend.json |     2 +-
 .../pmu-events/arch/x86/skylake/other.json    |     2 +-
 .../arch/x86/skylake/skl-metrics.json         |   178 +-
 .../arch/x86/skylake/uncore-cache.json        |   142 +
 .../arch/x86/skylake/uncore-other.json        |    79 +
 .../pmu-events/arch/x86/skylake/uncore.json   |   254 -
 .../arch/x86/skylake/virtual-memory.json      |     2 +-
 .../arch/x86/skylakex/floating-point.json     |     2 +-
 .../arch/x86/skylakex/frontend.json           |     2 +-
 .../pmu-events/arch/x86/skylakex/other.json   |    66 +-
 .../arch/x86/skylakex/pipeline.json           |    11 +
 .../arch/x86/skylakex/skx-metrics.json        |   667 +-
 .../arch/x86/skylakex/uncore-memory.json      |     9 +
 .../arch/x86/skylakex/uncore-other.json       |   730 +-
 .../arch/x86/skylakex/virtual-memory.json     |     2 +-
 .../x86/{tremontx => snowridgex}/cache.json   |    60 +-
 .../floating-point.json                       |     9 +-
 .../{tremontx => snowridgex}/frontend.json    |    20 +-
 .../x86/{tremontx => snowridgex}/memory.json  |     4 +-
 .../x86/{tremontx => snowridgex}/other.json   |    18 +-
 .../{tremontx => snowridgex}/pipeline.json    |    98 +-
 .../arch/x86/snowridgex/uncore-memory.json    |   619 +
 .../arch/x86/snowridgex/uncore-other.json     | 25249 ++++++++++++++++
 .../arch/x86/snowridgex/uncore-power.json     |   235 +
 .../virtual-memory.json                       |    69 +-
 .../pmu-events/arch/x86/tigerlake/cache.json  |    48 +-
 .../arch/x86/tigerlake/floating-point.json    |     2 +-
 .../arch/x86/tigerlake/frontend.json          |     2 +-
 .../pmu-events/arch/x86/tigerlake/memory.json |     2 +-
 .../pmu-events/arch/x86/tigerlake/other.json  |     1 -
 .../arch/x86/tigerlake/pipeline.json          |     4 +-
 .../arch/x86/tigerlake/tgl-metrics.json       |   378 +-
 .../arch/x86/tigerlake/uncore-other.json      |    65 +
 .../arch/x86/tigerlake/virtual-memory.json    |     2 +-
 .../arch/x86/tremontx/uncore-memory.json      |   245 -
 .../arch/x86/tremontx/uncore-other.json       |  2395 --
 .../arch/x86/tremontx/uncore-power.json       |    11 -
 .../arch/x86/westmereep-dp/cache.json         |     2 +-
 .../x86/westmereep-dp/floating-point.json     |     2 +-
 .../arch/x86/westmereep-dp/frontend.json      |     2 +-
 .../arch/x86/westmereep-dp/memory.json        |     2 +-
 .../x86/westmereep-dp/virtual-memory.json     |     2 +-
 .../x86/westmereep-sp/floating-point.json     |     2 +-
 .../arch/x86/westmereep-sp/frontend.json      |     2 +-
 .../x86/westmereep-sp/virtual-memory.json     |     2 +-
 .../arch/x86/westmereex/floating-point.json   |     2 +-
 .../arch/x86/westmereex/frontend.json         |     2 +-
 .../arch/x86/westmereex/virtual-memory.json   |     2 +-
 tools/perf/tests/pmu-events.c                 |     9 +
 252 files changed, 89144 insertions(+), 8438 deletions(-)
 create mode 100644 tools/perf/pmu-events/arch/x86/broadwell/uncore-cache.json
 create mode 100644 tools/perf/pmu-events/arch/x86/broadwell/uncore-other.json
 delete mode 100644 tools/perf/pmu-events/arch/x86/broadwell/uncore.json
 create mode 100644 tools/perf/pmu-events/arch/x86/broadwellde/uncore-other.json
 create mode 100644 tools/perf/pmu-events/arch/x86/broadwellx/uncore-other.json
 create mode 100644 tools/perf/pmu-events/arch/x86/haswellx/uncore-other.json
 create mode 100644 tools/perf/pmu-events/arch/x86/icelake/uncore-other.json
 create mode 100644 tools/perf/pmu-events/arch/x86/ivytown/uncore-other.json
 create mode 100644 tools/perf/pmu-events/arch/x86/jaketown/uncore-other.json
 delete mode 100644 tools/perf/pmu-events/arch/x86/knightslanding/uncore-memory.json
 create mode 100644 tools/perf/pmu-events/arch/x86/knightslanding/uncore-other.json
 create mode 100644 tools/perf/pmu-events/arch/x86/meteorlake/cache.json
 create mode 100644 tools/perf/pmu-events/arch/x86/meteorlake/frontend.json
 create mode 100644 tools/perf/pmu-events/arch/x86/meteorlake/memory.json
 create mode 100644 tools/perf/pmu-events/arch/x86/meteorlake/other.json
 create mode 100644 tools/perf/pmu-events/arch/x86/meteorlake/pipeline.json
 create mode 100644 tools/perf/pmu-events/arch/x86/meteorlake/virtual-memory.json
 create mode 100644 tools/perf/pmu-events/arch/x86/skylake/uncore-cache.json
 create mode 100644 tools/perf/pmu-events/arch/x86/skylake/uncore-other.json
 delete mode 100644 tools/perf/pmu-events/arch/x86/skylake/uncore.json
 rename tools/perf/pmu-events/arch/x86/{tremontx => snowridgex}/cache.json (95%)
 rename tools/perf/pmu-events/arch/x86/{tremontx => snowridgex}/floating-point.json (84%)
 rename tools/perf/pmu-events/arch/x86/{tremontx => snowridgex}/frontend.json (94%)
 rename tools/perf/pmu-events/arch/x86/{tremontx => snowridgex}/memory.json (99%)
 rename tools/perf/pmu-events/arch/x86/{tremontx => snowridgex}/other.json (98%)
 rename tools/perf/pmu-events/arch/x86/{tremontx => snowridgex}/pipeline.json (89%)
 create mode 100644 tools/perf/pmu-events/arch/x86/snowridgex/uncore-memory.json
 create mode 100644 tools/perf/pmu-events/arch/x86/snowridgex/uncore-other.json
 create mode 100644 tools/perf/pmu-events/arch/x86/snowridgex/uncore-power.json
 rename tools/perf/pmu-events/arch/x86/{tremontx => snowridgex}/virtual-memory.json (91%)
 create mode 100644 tools/perf/pmu-events/arch/x86/tigerlake/uncore-other.json
 delete mode 100644 tools/perf/pmu-events/arch/x86/tremontx/uncore-memory.json
 delete mode 100644 tools/perf/pmu-events/arch/x86/tremontx/uncore-other.json
 delete mode 100644 tools/perf/pmu-events/arch/x86/tremontx/uncore-power.json
Signature · 1 line
2.37.1.359.gd136c6c3e2-goog
author reviewer
#1 lore raw

PATCH 1/31 perf test: Avoid sysfs state affecting fake events

Icelake has a slots event, on my Skylakex I have CPU events in sysfs of topdown-slots-issued and topdown-total-slots. Legacy event parsing would try to use '-' to separate parts of an event and so perf_pmu__parse_init sets 'slots' to be a PMU_EVENT_SYMBOL_SUFFIX2. As such parsing the slots event for a fake PMU fails as a PMU_EVENT_SYMBOL_SUFFIX2 isn't made into the PE_PMU_EVENT_FAKE token. Resolve this issue by test initializing the PMU parsing state before every parse. This must be done every parse as the state is removes after each parse_events.

---
 tools/perf/tests/pmu-events.c | 9 +++++++++
 1 file changed, 9 insertions(+)
tools/perf/tests/pmu-events.c +9 -0
index 478b33825790..263cbb67c861 100644
@@ -812,6 +812,15 @@ static int check_parse_id(const char *id, struct parse_events_error *error,
 	for (cur = strchr(dup, '@') ; cur; cur = strchr(++cur, '@'))
 		*cur = '/';
 
+	if (fake_pmu) {
+		/*
+		 * Every call to __parse_events will try to initialize the PMU
+		 * state from sysfs and then clean it up at the end. Reset the
+		 * PMU events to the test state so that we don't pick up
+		 * erroneous prefixes and suffixes.
+		 */
+		perf_pmu__test_parse_init();
+	}
 	ret = __parse_events(evlist, dup, error, fake_pmu);
 	free(dup);
 
Signature · 1 line
2.37.1.359.gd136c6c3e2-goog
author reviewer
#2 lore raw

PATCH 2/31 perf vendor events: Update Intel broadwellx

In reply to Ian Rogers

Use script at:
https://github.com/intel/event-converter-for-linux-perf/blob/master/download_and_gen.py

to download and generate the latest events and metrics. Manually copy the broadwellx files into perf and update mapfile.csv.

Tested with 'perf test':

 10: PMU events                                                      :
 10.1: PMU event table sanity                                        : Ok
 10.2: PMU event map aliases                                         : Ok
 10.3: Parsing of PMU event table metrics                            : Ok
 10.4: Parsing of PMU event table metrics with fake PMUs             : Ok
 90: perf all metricgroups test                                      : Ok
 91: perf all metrics test                                           : Skip
 93: perf all PMU test                                               : Ok
---
 .../arch/x86/broadwellx/bdx-metrics.json      |  570 ++-
 .../pmu-events/arch/x86/broadwellx/cache.json |   22 +-
 .../arch/x86/broadwellx/floating-point.json   |    9 +-
 .../arch/x86/broadwellx/frontend.json         |    2 +-
 .../arch/x86/broadwellx/memory.json           |   39 +-
 .../pmu-events/arch/x86/broadwellx/other.json |    2 +-
 .../arch/x86/broadwellx/pipeline.json         |    4 +-
 .../arch/x86/broadwellx/uncore-cache.json     | 3788 ++++++++++++++++-
 .../x86/broadwellx/uncore-interconnect.json   | 1438 ++++++-
 .../arch/x86/broadwellx/uncore-memory.json    | 2849 ++++++++++++-
 .../arch/x86/broadwellx/uncore-other.json     | 3252 ++++++++++++++
 .../arch/x86/broadwellx/uncore-power.json     |  437 +-
 .../arch/x86/broadwellx/virtual-memory.json   |    2 +-
 tools/perf/pmu-events/arch/x86/mapfile.csv    |    2 +-
 14 files changed, 12103 insertions(+), 313 deletions(-)
 create mode 100644 tools/perf/pmu-events/arch/x86/broadwellx/uncore-other.json
tools/perf/pmu-events/arch/x86/broadwellx/bdx-metrics.json +517 -53
index b055947c0afe..720ee7c9332d 100644
@@ -74,12 +74,6 @@
         "MetricGroup": "Branches;Fed;FetchBW",
         "MetricName": "UpTB"
     },
-    {
-        "BriefDescription": "Cycles Per Instruction (per Logical Processor)",
-        "MetricExpr": "1 / (INST_RETIRED.ANY / CPU_CLK_UNHALTED.THREAD)",
-        "MetricGroup": "Pipeline;Mem",
-        "MetricName": "CPI"
-    },
     {
         "BriefDescription": "Per-Logical Processor actual clocks when the Logical Processor is active.",
         "MetricExpr": "CPU_CLK_UNHALTED.THREAD",
@@ -130,43 +124,25 @@
         "MetricName": "FLOPc_SMT"
     },
     {
-        "BriefDescription": "Actual per-core usage of the Floating Point execution units (regardless of the vector width)",
+        "BriefDescription": "Actual per-core usage of the Floating Point non-X87 execution units (regardless of precision or vector-width)",
         "MetricExpr": "( (FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE) + (FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE) ) / ( 2 * CPU_CLK_UNHALTED.THREAD )",
         "MetricGroup": "Cor;Flops;HPC",
         "MetricName": "FP_Arith_Utilization",
-        "PublicDescription": "Actual per-core usage of the Floating Point execution units (regardless of the vector width). Values > 1 are possible due to Fused-Multiply Add (FMA) counting."
+        "PublicDescription": "Actual per-core usage of the Floating Point non-X87 execution units (regardless of precision or vector-width). Values > 1 are possible due to ([BDW+] Fused-Multiply Add (FMA) counting - common; [ADL+] use all of ADD/MUL/FMA in Scalar or 128/256-bit vectors - less common)."
     },
     {
-        "BriefDescription": "Actual per-core usage of the Floating Point execution units (regardless of the vector width). SMT version; use when SMT is enabled and measuring per logical CPU.",
+        "BriefDescription": "Actual per-core usage of the Floating Point non-X87 execution units (regardless of precision or vector-width). SMT version; use when SMT is enabled and measuring per logical CPU.",
         "MetricExpr": "( (FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE) + (FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE) ) / ( 2 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ) )",
         "MetricGroup": "Cor;Flops;HPC_SMT",
         "MetricName": "FP_Arith_Utilization_SMT",
-        "PublicDescription": "Actual per-core usage of the Floating Point execution units (regardless of the vector width). Values > 1 are possible due to Fused-Multiply Add (FMA) counting. SMT version; use when SMT is enabled and measuring per logical CPU."
+        "PublicDescription": "Actual per-core usage of the Floating Point non-X87 execution units (regardless of precision or vector-width). Values > 1 are possible due to ([BDW+] Fused-Multiply Add (FMA) counting - common; [ADL+] use all of ADD/MUL/FMA in Scalar or 128/256-bit vectors - less common). SMT version; use when SMT is enabled and measuring per logical CPU."
     },
     {
-        "BriefDescription": "Instruction-Level-Parallelism (average number of uops executed when there is at least 1 uop executed)",
+        "BriefDescription": "Instruction-Level-Parallelism (average number of uops executed when there is execution) per-core",
         "MetricExpr": "UOPS_EXECUTED.THREAD / (( cpu@UOPS_EXECUTED.CORE\\,cmask\\=1@ / 2 ) if #SMT_on else UOPS_EXECUTED.CYCLES_GE_1_UOP_EXEC)",
         "MetricGroup": "Backend;Cor;Pipeline;PortsUtil",
         "MetricName": "ILP"
     },
-    {
-        "BriefDescription": "Branch Misprediction Cost: Fraction of TMA slots wasted per non-speculative branch misprediction (retired JEClear)",
-        "MetricExpr": " ( ((BR_MISP_RETIRED.ALL_BRANCHES / ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT )) * (( UOPS_ISSUED.ANY - UOPS_RETIRED.RETIRE_SLOTS + 4 * INT_MISC.RECOVERY_CYCLES ) / (4 * CPU_CLK_UNHALTED.THREAD))) + (4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) * (BR_MISP_RETIRED.ALL_BRANCHES * (12 * ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT + BACLEARS.ANY ) / CPU_CLK_UNHALTED.THREAD) / ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT + BACLEARS.ANY )) / #(4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) ) * (4 * CPU_CLK_UNHALTED.THREAD) / BR_MISP_RETIRED.ALL_BRANCHES",
-        "MetricGroup": "Bad;BrMispredicts",
-        "MetricName": "Branch_Misprediction_Cost"
-    },
-    {
-        "BriefDescription": "Branch Misprediction Cost: Fraction of TMA slots wasted per non-speculative branch misprediction (retired JEClear)",
-        "MetricExpr": " ( ((BR_MISP_RETIRED.ALL_BRANCHES / ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT )) * (( UOPS_ISSUED.ANY - UOPS_RETIRED.RETIRE_SLOTS + 4 * ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) ) / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )))) + (4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) * (BR_MISP_RETIRED.ALL_BRANCHES * (12 * ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT + BACLEARS.ANY ) / CPU_CLK_UNHALTED.THREAD) / ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT + BACLEARS.ANY )) / #(4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) ) * (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )) / BR_MISP_RETIRED.ALL_BRANCHES",
-        "MetricGroup": "Bad;BrMispredicts_SMT",
-        "MetricName": "Branch_Misprediction_Cost_SMT"
-    },
-    {
-        "BriefDescription": "Number of Instructions per non-speculative Branch Misprediction (JEClear)",
-        "MetricExpr": "INST_RETIRED.ANY / BR_MISP_RETIRED.ALL_BRANCHES",
-        "MetricGroup": "Bad;BadSpec;BrMispredicts",
-        "MetricName": "IpMispredict"
-    },
     {
         "BriefDescription": "Core actual clocks when any Logical Processor is active on the Physical Core",
         "MetricExpr": "( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )",
@@ -256,6 +232,18 @@
         "MetricGroup": "Summary;TmaL1",
         "MetricName": "Instructions"
     },
+    {
+        "BriefDescription": "Average number of Uops retired in cycles where at least one uop has retired.",
+        "MetricExpr": "UOPS_RETIRED.RETIRE_SLOTS / cpu@UOPS_RETIRED.RETIRE_SLOTS\\,cmask\\=1@",
+        "MetricGroup": "Pipeline;Ret",
+        "MetricName": "Retire"
+    },
+    {
+        "BriefDescription": "",
+        "MetricExpr": "UOPS_EXECUTED.THREAD / cpu@UOPS_EXECUTED.THREAD\\,cmask\\=1@",
+        "MetricGroup": "Cor;Pipeline;PortsUtil;SMT",
+        "MetricName": "Execute"
+    },
     {
         "BriefDescription": "Fraction of Uops delivered by the DSB (aka Decoded ICache; or Uop Cache)",
         "MetricExpr": "IDQ.DSB_UOPS / (( IDQ.DSB_UOPS + LSD.UOPS + IDQ.MITE_UOPS + IDQ.MS_UOPS ) )",
@@ -263,35 +251,34 @@
         "MetricName": "DSB_Coverage"
     },
     {
-        "BriefDescription": "Actual Average Latency for L1 data-cache miss demand load instructions (in core cycles)",
-        "MetricExpr": "L1D_PEND_MISS.PENDING / ( MEM_LOAD_UOPS_RETIRED.L1_MISS + mem_load_uops_retired.hit_lfb )",
-        "MetricGroup": "Mem;MemoryBound;MemoryLat",
-        "MetricName": "Load_Miss_Real_Latency",
-        "PublicDescription": "Actual Average Latency for L1 data-cache miss demand load instructions (in core cycles). Latency may be overestimated for multi-load instructions - e.g. repeat strings."
+        "BriefDescription": "Number of Instructions per non-speculative Branch Misprediction (JEClear) (lower number means higher occurrence rate)",
+        "MetricExpr": "INST_RETIRED.ANY / BR_MISP_RETIRED.ALL_BRANCHES",
+        "MetricGroup": "Bad;BadSpec;BrMispredicts",
+        "MetricName": "IpMispredict"
     },
     {
-        "BriefDescription": "Memory-Level-Parallelism (average number of L1 miss demand load when there is at least one such miss. Per-Logical Processor)",
-        "MetricExpr": "L1D_PEND_MISS.PENDING / L1D_PEND_MISS.PENDING_CYCLES",
-        "MetricGroup": "Mem;MemoryBound;MemoryBW",
-        "MetricName": "MLP"
+        "BriefDescription": "Branch Misprediction Cost: Fraction of TMA slots wasted per non-speculative branch misprediction (retired JEClear)",
+        "MetricExpr": " ( ((BR_MISP_RETIRED.ALL_BRANCHES / ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT )) * (( UOPS_ISSUED.ANY - UOPS_RETIRED.RETIRE_SLOTS + 4 * INT_MISC.RECOVERY_CYCLES ) / (4 * CPU_CLK_UNHALTED.THREAD))) + (4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) * (BR_MISP_RETIRED.ALL_BRANCHES * (12 * ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT + BACLEARS.ANY ) / CPU_CLK_UNHALTED.THREAD) / ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT + BACLEARS.ANY )) / #(4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) ) * (4 * CPU_CLK_UNHALTED.THREAD) / BR_MISP_RETIRED.ALL_BRANCHES",
+        "MetricGroup": "Bad;BrMispredicts",
+        "MetricName": "Branch_Misprediction_Cost"
     },
     {
-        "BriefDescription": "Average data fill bandwidth to the L1 data cache [GB / sec]",
-        "MetricExpr": "64 * L1D.REPLACEMENT / 1000000000 / duration_time",
-        "MetricGroup": "Mem;MemoryBW",
-        "MetricName": "L1D_Cache_Fill_BW"
+        "BriefDescription": "Branch Misprediction Cost: Fraction of TMA slots wasted per non-speculative branch misprediction (retired JEClear)",
+        "MetricExpr": " ( ((BR_MISP_RETIRED.ALL_BRANCHES / ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT )) * (( UOPS_ISSUED.ANY - UOPS_RETIRED.RETIRE_SLOTS + 4 * ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) ) / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )))) + (4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) * (BR_MISP_RETIRED.ALL_BRANCHES * (12 * ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT + BACLEARS.ANY ) / CPU_CLK_UNHALTED.THREAD) / ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT + BACLEARS.ANY )) / #(4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) ) * (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )) / BR_MISP_RETIRED.ALL_BRANCHES",
+        "MetricGroup": "Bad;BrMispredicts_SMT",
+        "MetricName": "Branch_Misprediction_Cost_SMT"
     },
     {
-        "BriefDescription": "Average data fill bandwidth to the L2 cache [GB / sec]",
-        "MetricExpr": "64 * L2_LINES_IN.ALL / 1000000000 / duration_time",
-        "MetricGroup": "Mem;MemoryBW",
-        "MetricName": "L2_Cache_Fill_BW"
+        "BriefDescription": "Actual Average Latency for L1 data-cache miss demand load operations (in core cycles)",
+        "MetricExpr": "L1D_PEND_MISS.PENDING / ( MEM_LOAD_UOPS_RETIRED.L1_MISS + mem_load_uops_retired.hit_lfb )",
+        "MetricGroup": "Mem;MemoryBound;MemoryLat",
+        "MetricName": "Load_Miss_Real_Latency"
     },
     {
-        "BriefDescription": "Average per-core data fill bandwidth to the L3 cache [GB / sec]",
-        "MetricExpr": "64 * LONGEST_LAT_CACHE.MISS / 1000000000 / duration_time",
-        "MetricGroup": "Mem;MemoryBW",
-        "MetricName": "L3_Cache_Fill_BW"
+        "BriefDescription": "Memory-Level-Parallelism (average number of L1 miss demand load when there is at least one such miss. Per-Logical Processor)",
+        "MetricExpr": "L1D_PEND_MISS.PENDING / L1D_PEND_MISS.PENDING_CYCLES",
+        "MetricGroup": "Mem;MemoryBound;MemoryBW",
+        "MetricName": "MLP"
     },
     {
         "BriefDescription": "L1 cache true misses per kilo instruction for retired demand loads",
@@ -306,13 +293,13 @@
         "MetricName": "L2MPKI"
     },
     {
-        "BriefDescription": "L2 cache misses per kilo instruction for all request types (including speculative)",
+        "BriefDescription": "L2 cache ([RKL+] true) misses per kilo instruction for all request types (including speculative)",
         "MetricExpr": "1000 * L2_RQSTS.MISS / INST_RETIRED.ANY",
         "MetricGroup": "Mem;CacheMisses;Offcore",
         "MetricName": "L2MPKI_All"
     },
     {
-        "BriefDescription": "L2 cache misses per kilo instruction for all demand loads  (including speculative)",
+        "BriefDescription": "L2 cache ([RKL+] true) misses per kilo instruction for all demand loads  (including speculative)",
         "MetricExpr": "1000 * L2_RQSTS.DEMAND_DATA_RD_MISS / INST_RETIRED.ANY",
         "MetricGroup": "Mem;CacheMisses",
         "MetricName": "L2MPKI_Load"
@@ -348,6 +335,48 @@
         "MetricGroup": "Mem;MemoryTLB_SMT",
         "MetricName": "Page_Walks_Utilization_SMT"
     },
+    {
+        "BriefDescription": "Average per-core data fill bandwidth to the L1 data cache [GB / sec]",
+        "MetricExpr": "64 * L1D.REPLACEMENT / 1000000000 / duration_time",
+        "MetricGroup": "Mem;MemoryBW",
+        "MetricName": "L1D_Cache_Fill_BW"
+    },
+    {
+        "BriefDescription": "Average per-core data fill bandwidth to the L2 cache [GB / sec]",
+        "MetricExpr": "64 * L2_LINES_IN.ALL / 1000000000 / duration_time",
+        "MetricGroup": "Mem;MemoryBW",
+        "MetricName": "L2_Cache_Fill_BW"
+    },
+    {
+        "BriefDescription": "Average per-core data fill bandwidth to the L3 cache [GB / sec]",
+        "MetricExpr": "64 * LONGEST_LAT_CACHE.MISS / 1000000000 / duration_time",
+        "MetricGroup": "Mem;MemoryBW",
+        "MetricName": "L3_Cache_Fill_BW"
+    },
+    {
+        "BriefDescription": "Average per-thread data fill bandwidth to the L1 data cache [GB / sec]",
+        "MetricExpr": "(64 * L1D.REPLACEMENT / 1000000000 / duration_time)",
+        "MetricGroup": "Mem;MemoryBW",
+        "MetricName": "L1D_Cache_Fill_BW_1T"
+    },
+    {
+        "BriefDescription": "Average per-thread data fill bandwidth to the L2 cache [GB / sec]",
+        "MetricExpr": "(64 * L2_LINES_IN.ALL / 1000000000 / duration_time)",
+        "MetricGroup": "Mem;MemoryBW",
+        "MetricName": "L2_Cache_Fill_BW_1T"
+    },
+    {
+        "BriefDescription": "Average per-thread data fill bandwidth to the L3 cache [GB / sec]",
+        "MetricExpr": "(64 * LONGEST_LAT_CACHE.MISS / 1000000000 / duration_time)",
+        "MetricGroup": "Mem;MemoryBW",
+        "MetricName": "L3_Cache_Fill_BW_1T"
+    },
+    {
+        "BriefDescription": "Average per-thread data access bandwidth to the L3 cache [GB / sec]",
+        "MetricExpr": "0",
+        "MetricGroup": "Mem;MemoryBW;Offcore",
+        "MetricName": "L3_Cache_Access_BW_1T"
+    },
     {
         "BriefDescription": "Average CPU Utilization",
         "MetricExpr": "CPU_CLK_UNHALTED.REF_TSC / msr@tsc@",
@@ -364,7 +393,8 @@
         "BriefDescription": "Giga Floating Point Operations Per Second",
         "MetricExpr": "( ( 1 * ( FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE ) + 2 * FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + 4 * ( FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE ) + 8 * FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE ) / 1000000000 ) / duration_time",
         "MetricGroup": "Cor;Flops;HPC",
-        "MetricName": "GFLOPs"
+        "MetricName": "GFLOPs",
+        "PublicDescription": "Giga Floating Point Operations Per Second. Aggregate across all supported options of: FP precisions, scalar and vector instructions, vector-width and AMX engine."
     },
     {
         "BriefDescription": "Average Frequency Utilization relative nominal frequency",
@@ -461,5 +491,439 @@
         "MetricExpr": "(cstate_pkg@c7\\-residency@ / msr@tsc@) * 100",
         "MetricGroup": "Power",
         "MetricName": "C7_Pkg_Residency"
+    },
+    {
+        "BriefDescription": "CPU operating frequency (in GHz)",
+        "MetricExpr": "( CPU_CLK_UNHALTED.THREAD / CPU_CLK_UNHALTED.REF_TSC * #SYSTEM_TSC_FREQ ) / 1000000000",
+        "MetricGroup": "",
+        "MetricName": "cpu_operating_frequency",
+        "ScaleUnit": "1GHz"
+    },
+    {
+        "BriefDescription": "Cycles per instruction retired; indicating how much time each executed instruction took; in units of cycles.",
+        "MetricExpr": "CPU_CLK_UNHALTED.THREAD / INST_RETIRED.ANY",
+        "MetricGroup": "",
+        "MetricName": "cpi",
+        "ScaleUnit": "1per_instr"
+    },
+    {
+        "BriefDescription": "The ratio of number of completed memory load instructions to the total number completed instructions",
+        "MetricExpr": "MEM_UOPS_RETIRED.ALL_LOADS / INST_RETIRED.ANY",
+        "MetricGroup": "",
+        "MetricName": "loads_per_instr",
+        "ScaleUnit": "1per_instr"
+    },
+    {
+        "BriefDescription": "The ratio of number of completed memory store instructions to the total number completed instructions",
+        "MetricExpr": "MEM_UOPS_RETIRED.ALL_STORES / INST_RETIRED.ANY",
+        "MetricGroup": "",
+        "MetricName": "stores_per_instr",
+        "ScaleUnit": "1per_instr"
+    },
+    {
+        "BriefDescription": "Ratio of number of requests missing L1 data cache (includes data+rfo w/ prefetches) to the total number of completed instructions",
+        "MetricExpr": "L1D.REPLACEMENT / INST_RETIRED.ANY",
+        "MetricGroup": "",
+        "MetricName": "l1d_mpi_includes_data_plus_rfo_with_prefetches",
+        "ScaleUnit": "1per_instr"
+    },
+    {
+        "BriefDescription": "Ratio of number of demand load requests hitting in L1 data cache to the total number of completed instructions",
+        "MetricExpr": "MEM_LOAD_UOPS_RETIRED.L1_HIT / INST_RETIRED.ANY",
+        "MetricGroup": "",
+        "MetricName": "l1d_demand_data_read_hits_per_instr",
+        "ScaleUnit": "1per_instr"
+    },
+    {
+        "BriefDescription": "Ratio of number of code read requests missing in L1 instruction cache (includes prefetches) to the total number of completed instructions",
+        "MetricExpr": "L2_RQSTS.ALL_CODE_RD / INST_RETIRED.ANY",
+        "MetricGroup": "",
+        "MetricName": "l1_i_code_read_misses_with_prefetches_per_instr",
+        "ScaleUnit": "1per_instr"
+    },
+    {
+        "BriefDescription": "Ratio of number of completed demand load requests hitting in L2 cache to the total number of completed instructions",
+        "MetricExpr": "MEM_LOAD_UOPS_RETIRED.L2_HIT / INST_RETIRED.ANY",
+        "MetricGroup": "",
+        "MetricName": "l2_demand_data_read_hits_per_instr",
+        "ScaleUnit": "1per_instr"
+    },
+    {
+        "BriefDescription": "Ratio of number of requests missing L2 cache (includes code+data+rfo w/ prefetches) to the total number of completed instructions",
+        "MetricExpr": "L2_LINES_IN.ALL / INST_RETIRED.ANY",
+        "MetricGroup": "",
+        "MetricName": "l2_mpi_includes_code_plus_data_plus_rfo_with_prefetches",
+        "ScaleUnit": "1per_instr"
+    },
+    {
+        "BriefDescription": "Ratio of number of completed data read request missing L2 cache to the total number of completed instructions",
+        "MetricExpr": "MEM_LOAD_UOPS_RETIRED.L2_MISS / INST_RETIRED.ANY",
+        "MetricGroup": "",
+        "MetricName": "l2_demand_data_read_mpi",
+        "ScaleUnit": "1per_instr"
+    },
+    {
+        "BriefDescription": "Ratio of number of code read request missing L2 cache to the total number of completed instructions",
+        "MetricExpr": "L2_RQSTS.CODE_RD_MISS / INST_RETIRED.ANY",
+        "MetricGroup": "",
+        "MetricName": "l2_demand_code_mpi",
+        "ScaleUnit": "1per_instr"
+    },
+    {
+        "BriefDescription": "Ratio of number of data read requests missing last level core cache (includes demand w/ prefetches) to the total number of completed instructions",
+        "MetricExpr": "( cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x182@ + cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x192@ ) / INST_RETIRED.ANY",
+        "MetricGroup": "",
+        "MetricName": "llc_data_read_mpi_demand_plus_prefetch",
+        "ScaleUnit": "1per_instr"
+    },
+    {
+        "BriefDescription": "Ratio of number of code read requests missing last level core cache (includes demand w/ prefetches) to the total number of completed instructions",
+        "MetricExpr": "( cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x181@ + cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x191@ ) / INST_RETIRED.ANY",
+        "MetricGroup": "",
+        "MetricName": "llc_code_read_mpi_demand_plus_prefetch",
+        "ScaleUnit": "1per_instr"
+    },
+    {
+        "BriefDescription": "Average latency of a last level cache (LLC) demand and prefetch data read miss (read memory access) in nano seconds",
+        "MetricExpr": "( 1000000000 * ( cbox@UNC_C_TOR_OCCUPANCY.MISS_OPCODE\\,filter_opc\\=0x182@ / cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x182@ ) / ( UNC_C_CLOCKTICKS / ( source_count(UNC_C_CLOCKTICKS) * #num_packages ) ) ) * duration_time",
+        "MetricGroup": "",
+        "MetricName": "llc_data_read_demand_plus_prefetch_miss_latency",
+        "ScaleUnit": "1ns"
+    },
+    {
+        "BriefDescription": "Average latency of a last level cache (LLC) demand and prefetch data read miss (read memory access) addressed to local memory in nano seconds",
+        "MetricExpr": "( 1000000000 * ( cbox@UNC_C_TOR_OCCUPANCY.MISS_LOCAL_OPCODE\\,filter_opc\\=0x182@ / cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x182@ ) / ( UNC_C_CLOCKTICKS / ( source_count(UNC_C_CLOCKTICKS) * #num_packages ) ) ) * duration_time",
+        "MetricGroup": "",
+        "MetricName": "llc_data_read_demand_plus_prefetch_miss_latency_for_local_requests",
+        "ScaleUnit": "1ns"
+    },
+    {
+        "BriefDescription": "Average latency of a last level cache (LLC) demand and prefetch data read miss (read memory access) addressed to remote memory in nano seconds",
+        "MetricExpr": "( 1000000000 * ( cbox@UNC_C_TOR_OCCUPANCY.MISS_REMOTE_OPCODE\\,filter_opc\\=0x182@ / cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x182@ ) / ( UNC_C_CLOCKTICKS / ( source_count(UNC_C_CLOCKTICKS) * #num_packages ) ) ) * duration_time",
+        "MetricGroup": "",
+        "MetricName": "llc_data_read_demand_plus_prefetch_miss_latency_for_remote_requests",
+        "ScaleUnit": "1ns"
+    },
+    {
+        "BriefDescription": "Ratio of number of completed page walks (for all page sizes) caused by a code fetch to the total number of completed instructions. This implies it missed in the ITLB (Instruction TLB) and further levels of TLB.",
+        "MetricExpr": "ITLB_MISSES.WALK_COMPLETED / INST_RETIRED.ANY",
+        "MetricGroup": "",
+        "MetricName": "itlb_mpi",
+        "ScaleUnit": "1per_instr"
+    },
+    {
+        "BriefDescription": "Ratio of number of completed page walks (for 2 megabyte and 4 megabyte page sizes) caused by a code fetch to the total number of completed instructions. This implies it missed in the Instruction Translation Lookaside Buffer (ITLB) and further levels of TLB.",
+        "MetricExpr": "ITLB_MISSES.WALK_COMPLETED_2M_4M / INST_RETIRED.ANY",
+        "MetricGroup": "",
+        "MetricName": "itlb_large_page_mpi",
+        "ScaleUnit": "1per_instr"
+    },
+    {
+        "BriefDescription": "Ratio of number of completed page walks (for all page sizes) caused by demand data loads to the total number of completed instructions. This implies it missed in the DTLB and further levels of TLB.",
+        "MetricExpr": "DTLB_LOAD_MISSES.WALK_COMPLETED / INST_RETIRED.ANY",
+        "MetricGroup": "",
+        "MetricName": "dtlb_load_mpi",
+        "ScaleUnit": "1per_instr"
+    },
+    {
+        "BriefDescription": "Ratio of number of completed page walks (for all page sizes) caused by demand data stores to the total number of completed instructions. This implies it missed in the DTLB and further levels of TLB.",
+        "MetricExpr": "DTLB_STORE_MISSES.WALK_COMPLETED / INST_RETIRED.ANY",
+        "MetricGroup": "",
+        "MetricName": "dtlb_store_mpi",
+        "ScaleUnit": "1per_instr"
+    },
+    {
+        "BriefDescription": "Memory read that miss the last level cache (LLC) addressed to local DRAM as a percentage of total memory read accesses, does not include LLC prefetches.",
+        "MetricExpr": "100 * cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x182@ / ( cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x182@ + cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x182@ )",
+        "MetricGroup": "",
+        "MetricName": "numa_percent_reads_addressed_to_local_dram",
+        "ScaleUnit": "1%"
+    },
+    {
+        "BriefDescription": "Memory reads that miss the last level cache (LLC) addressed to remote DRAM as a percentage of total memory read accesses, does not include LLC prefetches.",
+        "MetricExpr": "100 * cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x182@ / ( cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x182@ + cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x182@ )",
+        "MetricGroup": "",
+        "MetricName": "numa_percent_reads_addressed_to_remote_dram",
+        "ScaleUnit": "1%"
+    },
+    {
+        "BriefDescription": "Uncore operating frequency in GHz",
+        "MetricExpr": "UNC_C_CLOCKTICKS / ( source_count(UNC_C_CLOCKTICKS) * #num_packages ) / 1000000000",
+        "MetricGroup": "",
+        "MetricName": "uncore_frequency",
+        "ScaleUnit": "1GHz"
+    },
+    {
+        "BriefDescription": "Intel(R) Quick Path Interconnect (QPI) data transmit bandwidth (MB/sec)",
+        "MetricExpr": "( UNC_Q_TxL_FLITS_G0.DATA * 8 / 1000000) / duration_time",
+        "MetricGroup": "",
+        "MetricName": "qpi_data_transmit_bw_only_data",
+        "ScaleUnit": "1MB/s"
+    },
+    {
+        "BriefDescription": "DDR memory read bandwidth (MB/sec)",
+        "MetricExpr": "( UNC_M_CAS_COUNT.RD * 64 / 1000000) / duration_time",
+        "MetricGroup": "",
+        "MetricName": "memory_bandwidth_read",
+        "ScaleUnit": "1MB/s"
+    },
+    {
+        "BriefDescription": "DDR memory write bandwidth (MB/sec)",
+        "MetricExpr": "( UNC_M_CAS_COUNT.WR * 64 / 1000000) / duration_time",
+        "MetricGroup": "",
+        "MetricName": "memory_bandwidth_write",
+        "ScaleUnit": "1MB/s"
+    },
+    {
+        "BriefDescription": "DDR memory bandwidth (MB/sec)",
+        "MetricExpr": "(( UNC_M_CAS_COUNT.RD + UNC_M_CAS_COUNT.WR ) * 64 / 1000000) / duration_time",
+        "MetricGroup": "",
+        "MetricName": "memory_bandwidth_total",
+        "ScaleUnit": "1MB/s"
+    },
+    {
+        "BriefDescription": "Bandwidth of IO reads that are initiated by end device controllers that are requesting memory from the CPU.",
+        "MetricExpr": "( cbox@UNC_C_TOR_INSERTS.OPCODE\\,filter_opc\\=0x19e@ * 64 / 1000000) / duration_time",
+        "MetricGroup": "",
+        "MetricName": "io_bandwidth_read",
+        "ScaleUnit": "1MB/s"
+    },
+    {
+        "BriefDescription": "Bandwidth of IO writes that are initiated by end device controllers that are writing memory to the CPU.",
+        "MetricExpr": "(( cbox@UNC_C_TOR_INSERTS.OPCODE\\,filter_opc\\=0x1c8\\,filter_tid\\=0x3e@ + cbox@UNC_C_TOR_INSERTS.OPCODE\\,filter_opc\\=0x180\\,filter_tid\\=0x3e@ ) * 64 / 1000000) / duration_time",
+        "MetricGroup": "",
+        "MetricName": "io_bandwidth_write",
+        "ScaleUnit": "1MB/s"
+    },
+    {
+        "BriefDescription": "Uops delivered from decoded instruction cache (decoded stream buffer or DSB) as a percent of total uops delivered to Instruction Decode Queue",
+        "MetricExpr": "100 * ( IDQ.DSB_UOPS / UOPS_ISSUED.ANY )",
+        "MetricGroup": "",
+        "MetricName": "percent_uops_delivered_from_decoded_icache_dsb",
+        "ScaleUnit": "1%"
+    },
+    {
+        "BriefDescription": "Uops delivered from legacy decode pipeline (Micro-instruction Translation Engine or MITE) as a percent of total uops delivered to Instruction Decode Queue",
+        "MetricExpr": "100 * ( IDQ.MITE_UOPS / UOPS_ISSUED.ANY )",
+        "MetricGroup": "",
+        "MetricName": "percent_uops_delivered_from_legacy_decode_pipeline_mite",
+        "ScaleUnit": "1%"
+    },
+    {
+        "BriefDescription": "Uops delivered from microcode sequencer (MS) as a percent of total uops delivered to Instruction Decode Queue",
+        "MetricExpr": "100 * ( IDQ.MS_UOPS / UOPS_ISSUED.ANY )",
+        "MetricGroup": "",
+        "MetricName": "percent_uops_delivered_from_microcode_sequencer_ms",
+        "ScaleUnit": "1%"
+    },
+    {
+        "BriefDescription": "Uops delivered from loop stream detector(LSD) as a percent of total uops delivered to Instruction Decode Queue",
+        "MetricExpr": "100 * ( LSD.UOPS / UOPS_ISSUED.ANY )",
+        "MetricGroup": "",
+        "MetricName": "percent_uops_delivered_from_loop_stream_detector_lsd",
+        "ScaleUnit": "1%"
+    },
+    {
+        "BriefDescription": "This category represents fraction of slots where the processor's Frontend undersupplies its Backend. Frontend denotes the first part of the processor core responsible to fetch operations that are executed later on by the Backend part. Within the Frontend; a branch predictor predicts the next address to fetch; cache-lines are fetched from the memory subsystem; parsed into instructions; and lastly decoded into micro-operations (uops). Ideally the Frontend can issue Machine_Width uops every cycle to the Backend. Frontend Bound denotes unutilized issue-slots when there is no Backend stall; i.e. bubbles where Frontend delivered no uops while Backend could have accepted them. For example; stalls due to instruction-cache misses would be categorized under Frontend Bound.",
+        "MetricExpr": "100 * ( IDQ_UOPS_NOT_DELIVERED.CORE / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) )",
+        "MetricGroup": "TmaL1;PGO",
+        "MetricName": "tma_frontend_bound_percent",
+        "ScaleUnit": "1%"
+    },
+    {
+        "BriefDescription": "This metric represents fraction of slots the CPU was stalled due to Frontend latency issues.  For example; instruction-cache misses; iTLB misses or fetch stalls after a branch misprediction are categorized under Frontend Latency. In such cases; the Frontend eventually delivers no uops for some period.",
+        "MetricExpr": "100 * ( ( 4 ) * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) )",
+        "MetricGroup": "Frontend;TmaL2;m_tma_frontend_bound_percent",
+        "MetricName": "tma_fetch_latency_percent",
+        "ScaleUnit": "1%"
+    },
+    {
+        "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to instruction cache misses.",
+        "MetricExpr": "100 * ( ICACHE.IFDATA_STALL / ( CPU_CLK_UNHALTED.THREAD ) )",
+        "MetricGroup": "BigFoot;FetchLat;IcMiss;TmaL3;m_tma_fetch_latency_percent",
+        "MetricName": "tma_icache_misses_percent",
+        "ScaleUnit": "1%"
+    },
+    {
+        "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to Instruction TLB (ITLB) misses.",
+        "MetricExpr": "100 * ( ( 14 * ITLB_MISSES.STLB_HIT + cpu@ITLB_MISSES.WALK_DURATION\\,cmask\\=0x1@ + 7 * ITLB_MISSES.WALK_COMPLETED ) / ( CPU_CLK_UNHALTED.THREAD ) )",
+        "MetricGroup": "BigFoot;FetchLat;MemoryTLB;TmaL3;m_tma_fetch_latency_percent",
+        "MetricName": "tma_itlb_misses_percent",
+        "ScaleUnit": "1%"
+    },
+    {
+        "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to Branch Resteers. Branch Resteers estimates the Frontend delay in fetching operations from corrected path; following all sorts of miss-predicted branches. For example; branchy code with lots of miss-predictions might get categorized under Branch Resteers. Note the value of this node may overlap with its siblings.",
+        "MetricExpr": "100 * ( ( 12 ) * ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT + BACLEARS.ANY ) / ( CPU_CLK_UNHALTED.THREAD ) )",
+        "MetricGroup": "FetchLat;TmaL3;m_tma_fetch_latency_percent",
+        "MetricName": "tma_branch_resteers_percent",
+        "ScaleUnit": "1%"
+    },
+    {
+        "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to switches from DSB to MITE pipelines. The DSB (decoded i-cache) is a Uop Cache where the front-end directly delivers Uops (micro operations) avoiding heavy x86 decoding. The DSB pipeline has shorter latency and delivered higher bandwidth than the MITE (legacy instruction decode pipeline). Switching between the two pipelines can cause penalties hence this metric measures the exposed penalty.",
+        "MetricExpr": "100 * ( DSB2MITE_SWITCHES.PENALTY_CYCLES / ( CPU_CLK_UNHALTED.THREAD ) )",
+        "MetricGroup": "DSBmiss;FetchLat;TmaL3;m_tma_fetch_latency_percent",
+        "MetricName": "tma_dsb_switches_percent",
+        "ScaleUnit": "1%"
+    },
+    {
+        "BriefDescription": "This metric represents fraction of cycles CPU was stalled due to Length Changing Prefixes (LCPs). Using proper compiler flags or Intel Compiler by default will certainly avoid this. #Link: Optimization Guide about LCP BKMs.",
+        "MetricExpr": "100 * ( ILD_STALL.LCP / ( CPU_CLK_UNHALTED.THREAD ) )",
+        "MetricGroup": "FetchLat;TmaL3;m_tma_fetch_latency_percent",
+        "MetricName": "tma_lcp_percent",
+        "ScaleUnit": "1%"
+    },
+    {
+        "BriefDescription": "This metric estimates the fraction of cycles when the CPU was stalled due to switches of uop delivery to the Microcode Sequencer (MS). Commonly used instructions are optimized for delivery by the DSB (decoded i-cache) or MITE (legacy instruction decode) pipelines. Certain operations cannot be handled natively by the execution pipeline; and must be performed by microcode (small programs injected into the execution stream). Switching to the MS too often can negatively impact performance. The MS is designated to deliver long uop flows required by CISC instructions like CPUID; or uncommon conditions like Floating Point Assists when dealing with Denormals.",
+        "MetricExpr": "100 * ( ( 2 ) * IDQ.MS_SWITCHES / ( CPU_CLK_UNHALTED.THREAD ) )",
+        "MetricGroup": "FetchLat;MicroSeq;TmaL3;m_tma_fetch_latency_percent",
+        "MetricName": "tma_ms_switches_percent",
+        "ScaleUnit": "1%"
+    },
+    {
+        "BriefDescription": "This metric represents fraction of slots the CPU was stalled due to Frontend bandwidth issues.  For example; inefficiencies at the instruction decoders; or restrictions for caching in the DSB (decoded uops cache) are categorized under Fetch Bandwidth. In such cases; the Frontend typically delivers suboptimal amount of uops to the Backend.",
+        "MetricExpr": "100 * ( ( IDQ_UOPS_NOT_DELIVERED.CORE / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) - ( ( 4 ) * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) )",
+        "MetricGroup": "FetchBW;Frontend;TmaL2;m_tma_frontend_bound_percent",
+        "MetricName": "tma_fetch_bandwidth_percent",
+        "ScaleUnit": "1%"
+    },
+    {
+        "BriefDescription": "This metric represents Core fraction of cycles in which CPU was likely limited due to the MITE pipeline (the legacy decode pipeline). This pipeline is used for code that was not pre-cached in the DSB or LSD. For example; inefficiencies due to asymmetric decoders; use of long immediate or LCP can manifest as MITE fetch bandwidth bottleneck.",
+        "MetricExpr": "100 * ( ( IDQ.ALL_MITE_CYCLES_ANY_UOPS - IDQ.ALL_MITE_CYCLES_4_UOPS ) / ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) / 2 )",
+        "MetricGroup": "DSBmiss;FetchBW;TmaL3;m_tma_fetch_bandwidth_percent",
+        "MetricName": "tma_mite_percent",
+        "ScaleUnit": "1%"
+    },
+    {
+        "BriefDescription": "This metric represents Core fraction of cycles in which CPU was likely limited due to DSB (decoded uop cache) fetch pipeline.  For example; inefficient utilization of the DSB cache structure or bank conflict when reading from it; are categorized here.",
+        "MetricExpr": "100 * ( ( IDQ.ALL_DSB_CYCLES_ANY_UOPS - IDQ.ALL_DSB_CYCLES_4_UOPS ) / ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) / 2 )",
+        "MetricGroup": "DSB;FetchBW;TmaL3;m_tma_fetch_bandwidth_percent",
+        "MetricName": "tma_dsb_percent",
+        "ScaleUnit": "1%"
+    },
+    {
+        "BriefDescription": "This category represents fraction of slots wasted due to incorrect speculations. This include slots used to issue uops that do not eventually get retired and slots for which the issue-pipeline was blocked due to recovery from earlier incorrect speculation. For example; wasted work due to miss-predicted branches are categorized under Bad Speculation category. Incorrect data speculation followed by Memory Ordering Nukes is another example.",
+        "MetricExpr": "100 * ( ( UOPS_ISSUED.ANY - ( UOPS_RETIRED.RETIRE_SLOTS ) + ( 4 ) * ( ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) if #SMT_on else INT_MISC.RECOVERY_CYCLES ) ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) )",
+        "MetricGroup": "TmaL1",
+        "MetricName": "tma_bad_speculation_percent",
+        "ScaleUnit": "1%"
+    },
+    {
+        "BriefDescription": "This metric represents fraction of slots the CPU has wasted due to Branch Misprediction.  These slots are either wasted by uops fetched from an incorrectly speculated program path; or stalls when the out-of-order part of the machine needs to recover its state from a speculative path.",
+        "MetricExpr": "100 * ( ( BR_MISP_RETIRED.ALL_BRANCHES / ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT ) ) * ( ( UOPS_ISSUED.ANY - ( UOPS_RETIRED.RETIRE_SLOTS ) + ( 4 ) * ( ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) if #SMT_on else INT_MISC.RECOVERY_CYCLES ) ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) )",
+        "MetricGroup": "BadSpec;BrMispredicts;TmaL2;m_tma_bad_speculation_percent",
+        "MetricName": "tma_branch_mispredicts_percent",
+        "ScaleUnit": "1%"
+    },
+    {
+        "BriefDescription": "This metric represents fraction of slots the CPU has wasted due to Machine Clears.  These slots are either wasted by uops fetched prior to the clear; or stalls the out-of-order portion of the machine needs to recover its state after the clear. For example; this can happen due to memory ordering Nukes (e.g. Memory Disambiguation) or Self-Modifying-Code (SMC) nukes.",
+        "MetricExpr": "100 * ( ( ( UOPS_ISSUED.ANY - ( UOPS_RETIRED.RETIRE_SLOTS ) + ( 4 ) * ( ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) if #SMT_on else INT_MISC.RECOVERY_CYCLES ) ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) - ( ( BR_MISP_RETIRED.ALL_BRANCHES / ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT ) ) * ( ( UOPS_ISSUED.ANY - ( UOPS_RETIRED.RETIRE_SLOTS ) + ( 4 ) * ( ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) if #SMT_on else INT_MISC.RECOVERY_CYCLES ) ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) ) )",
+        "MetricGroup": "BadSpec;MachineClears;TmaL2;m_tma_bad_speculation_percent",
+        "MetricName": "tma_machine_clears_percent",
+        "ScaleUnit": "1%"
+    },
+    {
+        "BriefDescription": "This category represents fraction of slots where no uops are being delivered due to a lack of required resources for accepting new uops in the Backend. Backend is the portion of the processor core where the out-of-order scheduler dispatches ready uops into their respective execution units; and once completed these uops get retired according to program order. For example; stalls due to data-cache misses or stalls due to the divider unit being overloaded are both categorized under Backend Bound. Backend Bound is further divided into two main categories: Memory Bound and Core Bound.",
+        "MetricExpr": "100 * ( 1 - ( ( IDQ_UOPS_NOT_DELIVERED.CORE / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) + ( ( UOPS_ISSUED.ANY - ( UOPS_RETIRED.RETIRE_SLOTS ) + ( 4 ) * ( ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) if #SMT_on else INT_MISC.RECOVERY_CYCLES ) ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) + ( ( UOPS_RETIRED.RETIRE_SLOTS ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) ) )",
+        "MetricGroup": "TmaL1",
+        "MetricName": "tma_backend_bound_percent",
+        "ScaleUnit": "1%"
+    },
+    {
+        "BriefDescription": "This metric represents fraction of slots the Memory subsystem within the Backend was a bottleneck.  Memory Bound estimates fraction of slots where pipeline is likely stalled due to demand load or store instructions. This accounts mainly for (1) non-completed in-flight memory demand loads which coincides with execution units starvation; in addition to (2) cases where stores could impose backpressure on the pipeline when many of them get buffered at the same time (less common out of the two).",
+        "MetricExpr": "100 * ( ( ( CYCLE_ACTIVITY.STALLS_MEM_ANY + RESOURCE_STALLS.SB ) / ( ( CYCLE_ACTIVITY.STALLS_TOTAL + UOPS_EXECUTED.CYCLES_GE_1_UOP_EXEC - ( UOPS_EXECUTED.CYCLES_GE_3_UOPS_EXEC if ( ( INST_RETIRED.ANY / ( CPU_CLK_UNHALTED.THREAD ) ) > 1.8 ) else UOPS_EXECUTED.CYCLES_GE_2_UOPS_EXEC ) - ( RS_EVENTS.EMPTY_CYCLES if ( ( ( 4 ) * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) > 0.1 ) else 0 ) + RESOURCE_STALLS.SB ) ) ) * ( 1 - ( ( IDQ_UOPS_NOT_DELIVERED.CORE / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) + ( ( UOPS_ISSUED.ANY - ( UOPS_RETIRED.RETIRE_SLOTS ) + ( 4 ) * ( ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) if #SMT_on else INT_MISC.RECOVERY_CYCLES ) ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) + ( ( UOPS_RETIRED.RETIRE_SLOTS ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) ) ) )",
+        "MetricGroup": "Backend;TmaL2;m_tma_backend_bound_percent",
+        "MetricName": "tma_memory_bound_percent",
+        "ScaleUnit": "1%"
+    },
+    {
+        "BriefDescription": "This metric estimates how often the CPU was stalled without loads missing the L1 data cache.  The L1 data cache typically has the shortest latency.  However; in certain cases like loads blocked on older stores; a load might suffer due to high latency even though it is being satisfied by the L1. Another example is loads who miss in the TLB. These cases are characterized by execution unit stalls; while some non-completed demand load lives in the machine without having that demand load missing the L1 cache.",
+        "MetricExpr": "100 * ( max( ( CYCLE_ACTIVITY.STALLS_MEM_ANY - CYCLE_ACTIVITY.STALLS_L1D_MISS ) / ( CPU_CLK_UNHALTED.THREAD ) , 0 ) )",
+        "MetricGroup": "CacheMisses;MemoryBound;TmaL3mem;TmaL3;m_tma_memory_bound_percent",
+        "MetricName": "tma_l1_bound_percent",
+        "ScaleUnit": "1%"
+    },
+    {
+        "BriefDescription": "This metric estimates how often the CPU was stalled due to L2 cache accesses by loads.  Avoiding cache misses (i.e. L1 misses/L2 hits) can improve the latency and increase performance.",
+        "MetricExpr": "100 * ( ( CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS ) / ( CPU_CLK_UNHALTED.THREAD ) )",
+        "MetricGroup": "CacheMisses;MemoryBound;TmaL3mem;TmaL3;m_tma_memory_bound_percent",
+        "MetricName": "tma_l2_bound_percent",
+        "ScaleUnit": "1%"
+    },
+    {
+        "BriefDescription": "This metric estimates how often the CPU was stalled due to loads accesses to L3 cache or contended with a sibling Core.  Avoiding cache misses (i.e. L2 misses/L3 hits) can improve the latency and increase performance.",
+        "MetricExpr": "100 * ( ( MEM_LOAD_UOPS_RETIRED.L3_HIT / ( MEM_LOAD_UOPS_RETIRED.L3_HIT + ( 7 ) * MEM_LOAD_UOPS_RETIRED.L3_MISS ) ) * CYCLE_ACTIVITY.STALLS_L2_MISS / ( CPU_CLK_UNHALTED.THREAD ) )",
+        "MetricGroup": "CacheMisses;MemoryBound;TmaL3mem;TmaL3;m_tma_memory_bound_percent",
+        "MetricName": "tma_l3_bound_percent",
+        "ScaleUnit": "1%"
+    },
+    {
+        "BriefDescription": "This metric estimates how often the CPU was stalled on accesses to external memory (DRAM) by loads. Better caching can improve the latency and increase performance.",
+        "MetricExpr": "100 * ( min( ( ( 1 - ( MEM_LOAD_UOPS_RETIRED.L3_HIT / ( MEM_LOAD_UOPS_RETIRED.L3_HIT + ( 7 ) * MEM_LOAD_UOPS_RETIRED.L3_MISS ) ) ) * CYCLE_ACTIVITY.STALLS_L2_MISS / ( CPU_CLK_UNHALTED.THREAD ) ) , ( 1 ) ) )",
+        "MetricGroup": "MemoryBound;TmaL3mem;TmaL3;m_tma_memory_bound_percent",
+        "MetricName": "tma_dram_bound_percent",
+        "ScaleUnit": "1%"
+    },
+    {
+        "BriefDescription": "This metric estimates how often CPU was stalled  due to RFO store memory accesses; RFO store issue a read-for-ownership request before the write. Even though store accesses do not typically stall out-of-order CPUs; there are few cases where stores can lead to actual stalls. This metric will be flagged should RFO stores be a bottleneck.",
+        "MetricExpr": "100 * ( RESOURCE_STALLS.SB / ( CPU_CLK_UNHALTED.THREAD ) )",
+        "MetricGroup": "MemoryBound;TmaL3mem;TmaL3;m_tma_memory_bound_percent",
+        "MetricName": "tma_store_bound_percent",
+        "ScaleUnit": "1%"
+    },
+    {
+        "BriefDescription": "This metric represents fraction of slots where Core non-memory issues were of a bottleneck.  Shortage in hardware compute resources; or dependencies in software's instructions are both categorized under Core Bound. Hence it may indicate the machine ran out of an out-of-order resource; certain execution units are overloaded or dependencies in program's data- or instruction-flow are limiting the performance (e.g. FP-chained long-latency arithmetic operations).",
+        "MetricExpr": "100 * ( ( 1 - ( ( IDQ_UOPS_NOT_DELIVERED.CORE / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) + ( ( UOPS_ISSUED.ANY - ( UOPS_RETIRED.RETIRE_SLOTS ) + ( 4 ) * ( ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) if #SMT_on else INT_MISC.RECOVERY_CYCLES ) ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) + ( ( UOPS_RETIRED.RETIRE_SLOTS ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) ) ) - ( ( ( CYCLE_ACTIVITY.STALLS_MEM_ANY + RESOURCE_STALLS.SB ) / ( ( CYCLE_ACTIVITY.STALLS_TOTAL + UOPS_EXECUTED.CYCLES_GE_1_UOP_EXEC - ( UOPS_EXECUTED.CYCLES_GE_3_UOPS_EXEC if ( ( INST_RETIRED.ANY / ( CPU_CLK_UNHALTED.THREAD ) ) > 1.8 ) else UOPS_EXECUTED.CYCLES_GE_2_UOPS_EXEC ) - ( RS_EVENTS.EMPTY_CYCLES if ( ( ( 4 ) * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) > 0.1 ) else 0 ) + RESOURCE_STALLS.SB ) ) ) * ( 1 - ( ( IDQ_UOPS_NOT_DELIVERED.CORE / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) + ( ( UOPS_ISSUED.ANY - ( UOPS_RETIRED.RETIRE_SLOTS ) + ( 4 ) * ( ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) if #SMT_on else INT_MISC.RECOVERY_CYCLES ) ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) + ( ( UOPS_RETIRED.RETIRE_SLOTS ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) ) ) ) )",
+        "MetricGroup": "Backend;TmaL2;Compute;m_tma_backend_bound_percent",
+        "MetricName": "tma_core_bound_percent",
+        "ScaleUnit": "1%"
+    },
+    {
+        "BriefDescription": "This metric represents fraction of cycles where the Divider unit was active. Divide and square root instructions are performed by the Divider unit and can take considerably longer latency than integer or Floating Point addition; subtraction; or multiplication.",
+        "MetricExpr": "100 * ( ARITH.FPU_DIV_ACTIVE / ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) )",
+        "MetricGroup": "TmaL3;m_tma_core_bound_percent",
+        "MetricName": "tma_divider_percent",
+        "ScaleUnit": "1%"
+    },
+    {
+        "BriefDescription": "This metric estimates fraction of cycles the CPU performance was potentially limited due to Core computation issues (non divider-related).  Two distinct categories can be attributed into this metric: (1) heavy data-dependency among contiguous instructions would manifest in this metric - such cases are often referred to as low Instruction Level Parallelism (ILP). (2) Contention on some hardware execution unit other than Divider. For example; when there are too many multiply operations.",
+        "MetricExpr": "100 * ( ( ( ( CYCLE_ACTIVITY.STALLS_TOTAL + UOPS_EXECUTED.CYCLES_GE_1_UOP_EXEC - ( UOPS_EXECUTED.CYCLES_GE_3_UOPS_EXEC if ( ( INST_RETIRED.ANY / ( CPU_CLK_UNHALTED.THREAD ) ) > 1.8 ) else UOPS_EXECUTED.CYCLES_GE_2_UOPS_EXEC ) - ( RS_EVENTS.EMPTY_CYCLES if ( ( ( 4 ) * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) > 0.1 ) else 0 ) + RESOURCE_STALLS.SB ) ) - RESOURCE_STALLS.SB - CYCLE_ACTIVITY.STALLS_MEM_ANY ) / ( CPU_CLK_UNHALTED.THREAD ) )",
+        "MetricGroup": "PortsUtil;TmaL3;m_tma_core_bound_percent",
+        "MetricName": "tma_ports_utilization_percent",
+        "ScaleUnit": "1%"
+    },
+    {
+        "BriefDescription": "This category represents fraction of slots utilized by useful work i.e. issued uops that eventually get retired. Ideally; all pipeline slots would be attributed to the Retiring category.  Retiring of 100% would indicate the maximum Pipeline_Width throughput was achieved.  Maximizing Retiring typically increases the Instructions-per-cycle (see IPC metric). Note that a high Retiring value does not necessary mean there is no room for more performance.  For example; Heavy-operations or Microcode Assists are categorized under Retiring. They often indicate suboptimal performance and can often be optimized or avoided. ",
+        "MetricExpr": "100 * ( ( UOPS_RETIRED.RETIRE_SLOTS ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) )",
+        "MetricGroup": "TmaL1",
+        "MetricName": "tma_retiring_percent",
+        "ScaleUnit": "1%"
+    },
+    {
+        "BriefDescription": "This metric represents fraction of slots where the CPU was retiring light-weight operations -- instructions that require no more than one uop (micro-operation). This correlates with total number of instructions used by the program. A uops-per-instruction (see UPI metric) ratio of 1 or less should be expected for decently optimized software running on Intel Core/Xeon products. While this often indicates efficient X86 instructions were executed; high value does not necessarily mean better performance cannot be achieved.",
+        "MetricExpr": "100 * ( ( ( UOPS_RETIRED.RETIRE_SLOTS ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) - ( ( ( ( UOPS_RETIRED.RETIRE_SLOTS ) / UOPS_ISSUED.ANY ) * IDQ.MS_UOPS / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) ) )",
+        "MetricGroup": "Retire;TmaL2;m_tma_retiring_percent",
+        "MetricName": "tma_light_operations_percent",
+        "ScaleUnit": "1%"
+    },
+    {
+        "BriefDescription": "This metric represents overall arithmetic floating-point (FP) operations fraction the CPU has executed (retired). Note this metric's value may exceed its parent due to use of \"Uops\" CountDomain and FMA double-counting.",
+        "MetricExpr": "100 * ( ( INST_RETIRED.X87 * ( ( UOPS_RETIRED.RETIRE_SLOTS ) / INST_RETIRED.ANY ) / ( UOPS_RETIRED.RETIRE_SLOTS ) ) + ( ( FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE ) / ( UOPS_RETIRED.RETIRE_SLOTS ) ) + ( min( ( ( FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE ) / ( UOPS_RETIRED.RETIRE_SLOTS ) ) , ( 1 ) ) ) )",
+        "MetricGroup": "HPC;TmaL3;m_tma_light_operations_percent",
+        "MetricName": "tma_fp_arith_percent",
+        "ScaleUnit": "1%"
+    },
+    {
+        "BriefDescription": "This metric represents fraction of slots where the CPU was retiring heavy-weight operations -- instructions that require two or more uops or microcoded sequences. This highly-correlates with the uop length of these instructions/sequences.",
+        "MetricExpr": "100 * ( ( ( ( UOPS_RETIRED.RETIRE_SLOTS ) / UOPS_ISSUED.ANY ) * IDQ.MS_UOPS / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) )",
+        "MetricGroup": "Retire;TmaL2;m_tma_retiring_percent",
+        "MetricName": "tma_heavy_operations_percent",
+        "ScaleUnit": "1%"
+    },
+    {
+        "BriefDescription": "This metric represents fraction of slots the CPU was retiring uops fetched by the Microcode Sequencer (MS) unit.  The MS is used for CISC instructions not supported by the default decoders (like repeat move strings; or CPUID); or by microcode assists used to address some operation modes (like in Floating Point assists). These cases can often be avoided.",
+        "MetricExpr": "100 * ( ( ( UOPS_RETIRED.RETIRE_SLOTS ) / UOPS_ISSUED.ANY ) * IDQ.MS_UOPS / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) )",
+        "MetricGroup": "MicroSeq;TmaL3;m_tma_heavy_operations_percent",
+        "MetricName": "tma_microcode_sequencer_percent",
+        "ScaleUnit": "1%"
     }
 ]
tools/perf/pmu-events/arch/x86/broadwellx/cache.json +5 -17
index 127abe08362f..2efc4c0ee740 100644
@@ -814,9 +814,8 @@
         "EventCode": "0xB7, 0xBB",
         "EventName": "OFFCORE_RESPONSE.ALL_CODE_RD.LLC_HIT.HIT_OTHER_CORE_NO_FWD",
         "MSRIndex": "0x1a6,0x1a7",
-        "MSRValue": "0x04003C0244",
+        "MSRValue": "0x4003C0244",
         "Offcore": "1",
-        "PublicDescription": "Counts all demand & prefetch code reads hit in the L3 and the snoops to sibling cores hit in either E/S state and the line is not forwarded",
         "SampleAfterValue": "100003",
         "UMask": "0x1"
     },
@@ -829,7 +828,6 @@
         "MSRIndex": "0x1a6,0x1a7",
         "MSRValue": "0x10003C0091",
         "Offcore": "1",
-        "PublicDescription": "Counts all demand & prefetch data reads hit in the L3 and the snoop to one of the sibling cores hits the line in M state and the line is forwarded",
         "SampleAfterValue": "100003",
         "UMask": "0x1"
     },
@@ -840,9 +838,8 @@
         "EventCode": "0xB7, 0xBB",
         "EventName": "OFFCORE_RESPONSE.ALL_DATA_RD.LLC_HIT.HIT_OTHER_CORE_NO_FWD",
         "MSRIndex": "0x1a6,0x1a7",
-        "MSRValue": "0x04003C0091",
+        "MSRValue": "0x4003C0091",
         "Offcore": "1",
-        "PublicDescription": "Counts all demand & prefetch data reads hit in the L3 and the snoops to sibling cores hit in either E/S state and the line is not forwarded",
         "SampleAfterValue": "100003",
         "UMask": "0x1"
     },
@@ -855,7 +852,6 @@
         "MSRIndex": "0x1a6,0x1a7",
         "MSRValue": "0x10003C07F7",
         "Offcore": "1",
-        "PublicDescription": "Counts all data/code/rfo reads (demand & prefetch) hit in the L3 and the snoop to one of the sibling cores hits the line in M state and the line is forwarded",
         "SampleAfterValue": "100003",
         "UMask": "0x1"
     },
@@ -866,9 +862,8 @@
         "EventCode": "0xB7, 0xBB",
         "EventName": "OFFCORE_RESPONSE.ALL_READS.LLC_HIT.HIT_OTHER_CORE_NO_FWD",
         "MSRIndex": "0x1a6,0x1a7",
-        "MSRValue": "0x04003C07F7",
+        "MSRValue": "0x4003C07F7",
         "Offcore": "1",
-        "PublicDescription": "Counts all data/code/rfo reads (demand & prefetch) hit in the L3 and the snoops to sibling cores hit in either E/S state and the line is not forwarded",
         "SampleAfterValue": "100003",
         "UMask": "0x1"
     },
@@ -881,7 +876,6 @@
         "MSRIndex": "0x1a6,0x1a7",
         "MSRValue": "0x3F803C8FFF",
         "Offcore": "1",
-        "PublicDescription": "Counts all requests hit in the L3",
         "SampleAfterValue": "100003",
         "UMask": "0x1"
     },
@@ -894,7 +888,6 @@
         "MSRIndex": "0x1a6,0x1a7",
         "MSRValue": "0x10003C0122",
         "Offcore": "1",
-        "PublicDescription": "Counts all demand & prefetch RFOs hit in the L3 and the snoop to one of the sibling cores hits the line in M state and the line is forwarded",
         "SampleAfterValue": "100003",
         "UMask": "0x1"
     },
@@ -905,9 +898,8 @@
         "EventCode": "0xB7, 0xBB",
         "EventName": "OFFCORE_RESPONSE.ALL_RFO.LLC_HIT.HIT_OTHER_CORE_NO_FWD",
         "MSRIndex": "0x1a6,0x1a7",
-        "MSRValue": "0x04003C0122",
+        "MSRValue": "0x4003C0122",
         "Offcore": "1",
-        "PublicDescription": "Counts all demand & prefetch RFOs hit in the L3 and the snoops to sibling cores hit in either E/S state and the line is not forwarded",
         "SampleAfterValue": "100003",
         "UMask": "0x1"
     },
@@ -920,7 +912,6 @@
         "MSRIndex": "0x1a6,0x1a7",
         "MSRValue": "0x3F803C0002",
         "Offcore": "1",
-        "PublicDescription": "Counts all demand data writes (RFOs) hit in the L3",
         "SampleAfterValue": "100003",
         "UMask": "0x1"
     },
@@ -933,7 +924,6 @@
         "MSRIndex": "0x1a6,0x1a7",
         "MSRValue": "0x10003C0002",
         "Offcore": "1",
-        "PublicDescription": "Counts all demand data writes (RFOs) hit in the L3 and the snoop to one of the sibling cores hits the line in M state and the line is forwarded",
         "SampleAfterValue": "100003",
         "UMask": "0x1"
     },
@@ -946,7 +936,6 @@
         "MSRIndex": "0x1a6,0x1a7",
         "MSRValue": "0x3F803C0200",
         "Offcore": "1",
-        "PublicDescription": "Counts prefetch (that bring data to LLC only) code reads hit in the L3",
         "SampleAfterValue": "100003",
         "UMask": "0x1"
     },
@@ -959,7 +948,6 @@
         "MSRIndex": "0x1a6,0x1a7",
         "MSRValue": "0x3F803C0100",
         "Offcore": "1",
-        "PublicDescription": "Counts all prefetch (that bring data to LLC only) RFOs hit in the L3",
         "SampleAfterValue": "100003",
         "UMask": "0x1"
     },
@@ -973,4 +961,4 @@
         "SampleAfterValue": "100003",
         "UMask": "0x10"
     }
-]
\ No newline at end of file
+]
tools/perf/pmu-events/arch/x86/broadwellx/floating-point.json +8 -1
index 9ad37dddb354..93bbc8600321 100644
@@ -5,6 +5,7 @@
         "CounterHTOff": "0,1,2,3",
         "EventCode": "0xc7",
         "EventName": "FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE",
+        "PublicDescription": "Number of SSE/AVX computational 128-bit packed double precision floating-point instructions retired; some instructions will count twice as noted below.  Each count represents 2 computation operations, one for each element.  Applies to SSE* and AVX* packed double precision floating-point instructions: ADD SUB HADD HSUB SUBADD MUL DIV MIN MAX SQRT DPP FM(N)ADD/SUB.  DPP and FM(N)ADD/SUB instructions count twice as they perform 2 calculations per element. The DAZ and FTZ flags in the MXCSR register need to be set when using these events.",
         "SampleAfterValue": "2000003",
         "UMask": "0x4"
     },
@@ -14,6 +15,7 @@
         "CounterHTOff": "0,1,2,3",
         "EventCode": "0xc7",
         "EventName": "FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE",
+        "PublicDescription": "Number of SSE/AVX computational 128-bit packed single precision floating-point instructions retired; some instructions will count twice as noted below.  Each count represents 4 computation operations, one for each element.  Applies to SSE* and AVX* packed single precision floating-point instructions: ADD SUB HADD HSUB SUBADD MUL DIV MIN MAX SQRT RSQRT RCP DPP FM(N)ADD/SUB.  DPP and FM(N)ADD/SUB instructions count twice as they perform 2 calculations per element. The DAZ and FTZ flags in the MXCSR register need to be set when using these events.",
         "SampleAfterValue": "2000003",
         "UMask": "0x8"
     },
@@ -23,6 +25,7 @@
         "CounterHTOff": "0,1,2,3",
         "EventCode": "0xc7",
         "EventName": "FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE",
+        "PublicDescription": "Number of SSE/AVX computational 256-bit packed double precision floating-point instructions retired; some instructions will count twice as noted below.  Each count represents 4 computation operations, one for each element.  Applies to SSE* and AVX* packed double precision floating-point instructions: ADD SUB HADD HSUB SUBADD MUL DIV MIN MAX SQRT FM(N)ADD/SUB.  FM(N)ADD/SUB instructions count twice as they perform 2 calculations per element. The DAZ and FTZ flags in the MXCSR register need to be set when using these events.",
         "SampleAfterValue": "2000003",
         "UMask": "0x10"
     },
@@ -32,6 +35,7 @@
         "CounterHTOff": "0,1,2,3",
         "EventCode": "0xc7",
         "EventName": "FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE",
+        "PublicDescription": "Number of SSE/AVX computational 256-bit packed single precision floating-point instructions retired; some instructions will count twice as noted below.  Each count represents 8 computation operations, one for each element.  Applies to SSE* and AVX* packed single precision floating-point instructions: ADD SUB HADD HSUB SUBADD MUL DIV MIN MAX SQRT RSQRT RCP DPP FM(N)ADD/SUB.  DPP and FM(N)ADD/SUB instructions count twice as they perform 2 calculations per element. The DAZ and FTZ flags in the MXCSR register need to be set when using these events.",
         "SampleAfterValue": "2000003",
         "UMask": "0x20"
     },
@@ -59,6 +63,7 @@
         "CounterHTOff": "0,1,2,3",
         "EventCode": "0xc7",
         "EventName": "FP_ARITH_INST_RETIRED.SCALAR",
+        "PublicDescription": "Number of SSE/AVX computational scalar single precision and double precision floating-point instructions retired; some instructions will count twice as noted below.  Each count represents 1 computational operation. Applies to SSE* and AVX* scalar single precision floating-point instructions: ADD SUB MUL DIV MIN MAX SQRT RSQRT RCP FM(N)ADD/SUB.  FM(N)ADD/SUB instructions count twice as they perform 2 calculations per element. The DAZ and FTZ flags in the MXCSR register need to be set when using these events.",
         "SampleAfterValue": "2000003",
         "UMask": "0x3"
     },
@@ -68,6 +73,7 @@
         "CounterHTOff": "0,1,2,3",
         "EventCode": "0xc7",
         "EventName": "FP_ARITH_INST_RETIRED.SCALAR_DOUBLE",
+        "PublicDescription": "Number of SSE/AVX computational scalar double precision floating-point instructions retired; some instructions will count twice as noted below.  Each count represents 1 computational operation. Applies to SSE* and AVX* scalar double precision floating-point instructions: ADD SUB MUL DIV MIN MAX SQRT FM(N)ADD/SUB.  FM(N)ADD/SUB instructions count twice as they perform 2 calculations per element. The DAZ and FTZ flags in the MXCSR register need to be set when using these events.",
         "SampleAfterValue": "2000003",
         "UMask": "0x1"
     },
@@ -77,6 +83,7 @@
         "CounterHTOff": "0,1,2,3",
         "EventCode": "0xc7",
         "EventName": "FP_ARITH_INST_RETIRED.SCALAR_SINGLE",
+        "PublicDescription": "Number of SSE/AVX computational scalar single precision floating-point instructions retired; some instructions will count twice as noted below.  Each count represents 1 computational operation. Applies to SSE* and AVX* scalar single precision floating-point instructions: ADD SUB MUL DIV MIN MAX SQRT RSQRT RCP FM(N)ADD/SUB.  FM(N)ADD/SUB instructions count twice as they perform 2 calculations per element. The DAZ and FTZ flags in the MXCSR register need to be set when using these events.",
         "SampleAfterValue": "2000003",
         "UMask": "0x2"
     },
@@ -190,4 +197,4 @@
         "SampleAfterValue": "2000003",
         "UMask": "0x3"
     }
-]
\ No newline at end of file
+]
tools/perf/pmu-events/arch/x86/broadwellx/frontend.json +1 -1
index f0bcb945ff76..37ce8034b2ed 100644
@@ -292,4 +292,4 @@
         "SampleAfterValue": "2000003",
         "UMask": "0x1"
     }
-]
\ No newline at end of file
+]
tools/perf/pmu-events/arch/x86/broadwellx/memory.json +10 -29
index cce993b197e3..545f61f691b9 100644
@@ -247,7 +247,6 @@
         "MSRIndex": "0x1a6,0x1a7",
         "MSRValue": "0x3FBFC00244",
         "Offcore": "1",
-        "PublicDescription": "Counts all demand & prefetch code reads miss in the L3",
         "SampleAfterValue": "100003",
         "UMask": "0x1"
     },
@@ -258,9 +257,8 @@
         "EventCode": "0xB7, 0xBB",
         "EventName": "OFFCORE_RESPONSE.ALL_CODE_RD.LLC_MISS.LOCAL_DRAM",
         "MSRIndex": "0x1a6,0x1a7",
-        "MSRValue": "0x0604000244",
+        "MSRValue": "0x604000244",
         "Offcore": "1",
-        "PublicDescription": "Counts all demand & prefetch code reads miss the L3 and the data is returned from local dram",
         "SampleAfterValue": "100003",
         "UMask": "0x1"
     },
@@ -273,7 +271,6 @@
         "MSRIndex": "0x1a6,0x1a7",
         "MSRValue": "0x3FBFC00091",
         "Offcore": "1",
-        "PublicDescription": "Counts all demand & prefetch data reads miss in the L3",
         "SampleAfterValue": "100003",
         "UMask": "0x1"
     },
@@ -284,9 +281,8 @@
         "EventCode": "0xB7, 0xBB",
         "EventName": "OFFCORE_RESPONSE.ALL_DATA_RD.LLC_MISS.LOCAL_DRAM",
         "MSRIndex": "0x1a6,0x1a7",
-        "MSRValue": "0x0604000091",
+        "MSRValue": "0x604000091",
         "Offcore": "1",
-        "PublicDescription": "Counts all demand & prefetch data reads miss the L3 and the data is returned from local dram",
         "SampleAfterValue": "100003",
         "UMask": "0x1"
     },
@@ -297,9 +293,8 @@
         "EventCode": "0xB7, 0xBB",
         "EventName": "OFFCORE_RESPONSE.ALL_DATA_RD.LLC_MISS.REMOTE_DRAM",
         "MSRIndex": "0x1a6,0x1a7",
-        "MSRValue": "0x063BC00091",
+        "MSRValue": "0x63BC00091",
         "Offcore": "1",
-        "PublicDescription": "Counts all demand & prefetch data reads miss the L3 and the data is returned from remote dram",
         "SampleAfterValue": "100003",
         "UMask": "0x1"
     },
@@ -312,7 +307,6 @@
         "MSRIndex": "0x1a6,0x1a7",
         "MSRValue": "0x103FC00091",
         "Offcore": "1",
-        "PublicDescription": "Counts all demand & prefetch data reads miss the L3 and the modified data is transferred from remote cache",
         "SampleAfterValue": "100003",
         "UMask": "0x1"
     },
@@ -323,9 +317,8 @@
         "EventCode": "0xB7, 0xBB",
         "EventName": "OFFCORE_RESPONSE.ALL_DATA_RD.LLC_MISS.REMOTE_HIT_FORWARD",
         "MSRIndex": "0x1a6,0x1a7",
-        "MSRValue": "0x087FC00091",
+        "MSRValue": "0x87FC00091",
         "Offcore": "1",
-        "PublicDescription": "Counts all demand & prefetch data reads miss the L3 and clean or shared data is transferred from remote cache",
         "SampleAfterValue": "100003",
         "UMask": "0x1"
     },
@@ -338,20 +331,18 @@
         "MSRIndex": "0x1a6,0x1a7",
         "MSRValue": "0x3FBFC007F7",
         "Offcore": "1",
-        "PublicDescription": "Counts all data/code/rfo reads (demand & prefetch) miss in the L3",
         "SampleAfterValue": "100003",
         "UMask": "0x1"
     },
     {
-        "BriefDescription": "Counts all data/code/rfo reads (demand & prefetch)miss the L3 and the data is returned from local dram",
+        "BriefDescription": "Counts all data/code/rfo reads (demand & prefetch) miss the L3 and the data is returned from local dram",
         "Counter": "0,1,2,3",
         "CounterHTOff": "0,1,2,3",
         "EventCode": "0xB7, 0xBB",
         "EventName": "OFFCORE_RESPONSE.ALL_READS.LLC_MISS.LOCAL_DRAM",
         "MSRIndex": "0x1a6,0x1a7",
-        "MSRValue": "0x06040007F7",
+        "MSRValue": "0x6040007F7",
         "Offcore": "1",
-        "PublicDescription": "Counts all data/code/rfo reads (demand & prefetch)miss the L3 and the data is returned from local dram",
         "SampleAfterValue": "100003",
         "UMask": "0x1"
     },
@@ -362,9 +353,8 @@
         "EventCode": "0xB7, 0xBB",
         "EventName": "OFFCORE_RESPONSE.ALL_READS.LLC_MISS.REMOTE_DRAM",
         "MSRIndex": "0x1a6,0x1a7",
-        "MSRValue": "0x063BC007F7",
+        "MSRValue": "0x63BC007F7",
         "Offcore": "1",
-        "PublicDescription": "Counts all data/code/rfo reads (demand & prefetch) miss the L3 and the data is returned from remote dram",
         "SampleAfterValue": "100003",
         "UMask": "0x1"
     },
@@ -377,7 +367,6 @@
         "MSRIndex": "0x1a6,0x1a7",
         "MSRValue": "0x103FC007F7",
         "Offcore": "1",
-        "PublicDescription": "Counts all data/code/rfo reads (demand & prefetch) miss the L3 and the modified data is transferred from remote cache",
         "SampleAfterValue": "100003",
         "UMask": "0x1"
     },
@@ -388,9 +377,8 @@
         "EventCode": "0xB7, 0xBB",
         "EventName": "OFFCORE_RESPONSE.ALL_READS.LLC_MISS.REMOTE_HIT_FORWARD",
         "MSRIndex": "0x1a6,0x1a7",
-        "MSRValue": "0x087FC007F7",
+        "MSRValue": "0x87FC007F7",
         "Offcore": "1",
-        "PublicDescription": "Counts all data/code/rfo reads (demand & prefetch) miss the L3 and clean or shared data is transferred from remote cache",
         "SampleAfterValue": "100003",
         "UMask": "0x1"
     },
@@ -403,7 +391,6 @@
         "MSRIndex": "0x1a6,0x1a7",
         "MSRValue": "0x3FBFC08FFF",
         "Offcore": "1",
-        "PublicDescription": "Counts all requests miss in the L3",
         "SampleAfterValue": "100003",
         "UMask": "0x1"
     },
@@ -416,7 +403,6 @@
         "MSRIndex": "0x1a6,0x1a7",
         "MSRValue": "0x3FBFC00122",
         "Offcore": "1",
-        "PublicDescription": "Counts all demand & prefetch RFOs miss in the L3",
         "SampleAfterValue": "100003",
         "UMask": "0x1"
     },
@@ -427,9 +413,8 @@
         "EventCode": "0xB7, 0xBB",
         "EventName": "OFFCORE_RESPONSE.ALL_RFO.LLC_MISS.LOCAL_DRAM",
         "MSRIndex": "0x1a6,0x1a7",
-        "MSRValue": "0x0604000122",
+        "MSRValue": "0x604000122",
         "Offcore": "1",
-        "PublicDescription": "Counts all demand & prefetch RFOs miss the L3 and the data is returned from local dram",
         "SampleAfterValue": "100003",
         "UMask": "0x1"
     },
@@ -442,7 +427,6 @@
         "MSRIndex": "0x1a6,0x1a7",
         "MSRValue": "0x3FBFC00002",
         "Offcore": "1",
-        "PublicDescription": "Counts all demand data writes (RFOs) miss in the L3",
         "SampleAfterValue": "100003",
         "UMask": "0x1"
     },
@@ -455,7 +439,6 @@
         "MSRIndex": "0x1a6,0x1a7",
         "MSRValue": "0x103FC00002",
         "Offcore": "1",
-        "PublicDescription": "Counts all demand data writes (RFOs) miss the L3 and the modified data is transferred from remote cache",
         "SampleAfterValue": "100003",
         "UMask": "0x1"
     },
@@ -468,7 +451,6 @@
         "MSRIndex": "0x1a6,0x1a7",
         "MSRValue": "0x3FBFC00200",
         "Offcore": "1",
-        "PublicDescription": "Counts prefetch (that bring data to LLC only) code reads miss in the L3",
         "SampleAfterValue": "100003",
         "UMask": "0x1"
     },
@@ -481,7 +463,6 @@
         "MSRIndex": "0x1a6,0x1a7",
         "MSRValue": "0x3FBFC00100",
         "Offcore": "1",
-        "PublicDescription": "Counts all prefetch (that bring data to LLC only) RFOs miss in the L3",
         "SampleAfterValue": "100003",
         "UMask": "0x1"
     },
@@ -684,4 +665,4 @@
         "SampleAfterValue": "2000003",
         "UMask": "0x40"
     }
-]
\ No newline at end of file
+]
tools/perf/pmu-events/arch/x86/broadwellx/other.json +1 -1
index 4b360fe96698..917d145d5227 100644
@@ -41,4 +41,4 @@
         "SampleAfterValue": "2000003",
         "UMask": "0x1"
     }
-]
\ No newline at end of file
+]
tools/perf/pmu-events/arch/x86/broadwellx/pipeline.json +2 -2
index 18d21b94a4b9..f0f30081d683 100644
@@ -1369,7 +1369,7 @@
         "BriefDescription": "Cycles with less than 10 actually retired uops.",
         "Counter": "0,1,2,3",
         "CounterHTOff": "0,1,2,3",
-        "CounterMask": "10",
+        "CounterMask": "16",
         "EventCode": "0xC2",
         "EventName": "UOPS_RETIRED.TOTAL_CYCLES",
         "Invert": "1",
@@ -1377,4 +1377,4 @@
         "SampleAfterValue": "2000003",
         "UMask": "0x1"
     }
-]
\ No newline at end of file
+]
tools/perf/pmu-events/arch/x86/broadwellx/uncore-cache.json +3637 -151

2,005 lines not on this page · raw message

tools/perf/pmu-events/arch/x86/broadwellx/uncore-interconnect.json +1432 -6

1,438 lines not on this page · raw message

tools/perf/pmu-events/arch/x86/broadwellx/uncore-memory.json +2835 -14

2,849 lines not on this page · raw message

tools/perf/pmu-events/arch/x86/broadwellx/uncore-other.json new file +3252 -0

3,252 lines not on this page · raw message

tools/perf/pmu-events/arch/x86/broadwellx/uncore-power.json +401 -36

437 lines not on this page · raw message

tools/perf/pmu-events/arch/x86/broadwellx/virtual-memory.json +1 -1
index 818a8b132c08..6a6de8790f25 100644

Lines of this file are not shown. View the raw message.

tools/perf/pmu-events/arch/x86/mapfile.csv +1 -1
index f5a382421a60..02d3434aecf3 100644

Lines of this file are not shown. View the raw message.

The rest of this diff is not shown. View the raw message.

Signature · 1 line
2.37.1.359.gd136c6c3e2-goog

Part of the diff is not shown here. View the raw message.

#3 lore raw

Re: [PATCH v1 02/31] perf vendor events: Update Intel broadwellx

Hi Ian,

Quoted 3 lines · Ian Rogers

Use script at:
https://github.com/intel/event-converter-for-linux-perf/blob/master/download_and_gen.py

It is better to add the event JSON file version and TMA version for the future track. For example, the event list is based on broadwellx JSON file v19, the metrics are based on TMA4.4-full.

Quoted 13,219 lines

to download and generate the latest events and metrics. Manually copy the broadwellx files into perf and update mapfile.csv.

Tested with 'perf test':

  10: PMU events                                                      :
  10.1: PMU event table sanity                                        : Ok
  10.2: PMU event map aliases                                         : Ok
  10.3: Parsing of PMU event table metrics                            : Ok
  10.4: Parsing of PMU event table metrics with fake PMUs             : Ok
  90: perf all metricgroups test                                      : Ok
  91: perf all metrics test                                           : Skip
  93: perf all PMU test                                               : Ok
---
  .../arch/x86/broadwellx/bdx-metrics.json      |  570 ++-
  .../pmu-events/arch/x86/broadwellx/cache.json |   22 +-
  .../arch/x86/broadwellx/floating-point.json   |    9 +-
  .../arch/x86/broadwellx/frontend.json         |    2 +-
  .../arch/x86/broadwellx/memory.json           |   39 +-
  .../pmu-events/arch/x86/broadwellx/other.json |    2 +-
  .../arch/x86/broadwellx/pipeline.json         |    4 +-
  .../arch/x86/broadwellx/uncore-cache.json     | 3788 ++++++++++++++++-
  .../x86/broadwellx/uncore-interconnect.json   | 1438 ++++++-
  .../arch/x86/broadwellx/uncore-memory.json    | 2849 ++++++++++++-
  .../arch/x86/broadwellx/uncore-other.json     | 3252 ++++++++++++++
  .../arch/x86/broadwellx/uncore-power.json     |  437 +-
  .../arch/x86/broadwellx/virtual-memory.json   |    2 +-
  tools/perf/pmu-events/arch/x86/mapfile.csv    |    2 +-
  14 files changed, 12103 insertions(+), 313 deletions(-)

create mode 100644 tools/perf/pmu-events/arch/x86/broadwellx/uncore-other.json

tools/perf/pmu-events/arch/x86/broadwellx/bdx-metrics.json +517 -53
index b055947c0afe..720ee7c9332d 100644
@@ -74,12 +74,6 @@
          "MetricGroup": "Branches;Fed;FetchBW",
          "MetricName": "UpTB"
      },
-    {
-        "BriefDescription": "Cycles Per Instruction (per Logical Processor)",
-        "MetricExpr": "1 / (INST_RETIRED.ANY / CPU_CLK_UNHALTED.THREAD)",
-        "MetricGroup": "Pipeline;Mem",
-        "MetricName": "CPI"
-    },
      {
          "BriefDescription": "Per-Logical Processor actual clocks when the Logical Processor is active.",
          "MetricExpr": "CPU_CLK_UNHALTED.THREAD",
@@ -130,43 +124,25 @@
          "MetricName": "FLOPc_SMT"
      },
      {
-        "BriefDescription": "Actual per-core usage of the Floating Point execution units (regardless of the vector width)",
+        "BriefDescription": "Actual per-core usage of the Floating Point non-X87 execution units (regardless of precision or vector-width)",
          "MetricExpr": "( (FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE) + (FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE) ) / ( 2 * CPU_CLK_UNHALTED.THREAD )",
          "MetricGroup": "Cor;Flops;HPC",
          "MetricName": "FP_Arith_Utilization",
-        "PublicDescription": "Actual per-core usage of the Floating Point execution units (regardless of the vector width). Values > 1 are possible due to Fused-Multiply Add (FMA) counting."
+        "PublicDescription": "Actual per-core usage of the Floating Point non-X87 execution units (regardless of precision or vector-width). Values > 1 are possible due to ([BDW+] Fused-Multiply Add (FMA) counting - common; [ADL+] use all of ADD/MUL/FMA in Scalar or 128/256-bit vectors - less common)."
      },
      {
-        "BriefDescription": "Actual per-core usage of the Floating Point execution units (regardless of the vector width). SMT version; use when SMT is enabled and measuring per logical CPU.",
+        "BriefDescription": "Actual per-core usage of the Floating Point non-X87 execution units (regardless of precision or vector-width). SMT version; use when SMT is enabled and measuring per logical CPU.",
          "MetricExpr": "( (FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE) + (FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE) ) / ( 2 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ) )",
          "MetricGroup": "Cor;Flops;HPC_SMT",
          "MetricName": "FP_Arith_Utilization_SMT",
-        "PublicDescription": "Actual per-core usage of the Floating Point execution units (regardless of the vector width). Values > 1 are possible due to Fused-Multiply Add (FMA) counting. SMT version; use when SMT is enabled and measuring per logical CPU."
+        "PublicDescription": "Actual per-core usage of the Floating Point non-X87 execution units (regardless of precision or vector-width). Values > 1 are possible due to ([BDW+] Fused-Multiply Add (FMA) counting - common; [ADL+] use all of ADD/MUL/FMA in Scalar or 128/256-bit vectors - less common). SMT version; use when SMT is enabled and measuring per logical CPU."
      },
      {
-        "BriefDescription": "Instruction-Level-Parallelism (average number of uops executed when there is at least 1 uop executed)",
+        "BriefDescription": "Instruction-Level-Parallelism (average number of uops executed when there is execution) per-core",
          "MetricExpr": "UOPS_EXECUTED.THREAD / (( cpu@UOPS_EXECUTED.CORE\\,cmask\\=1@ / 2 ) if #SMT_on else UOPS_EXECUTED.CYCLES_GE_1_UOP_EXEC)",
          "MetricGroup": "Backend;Cor;Pipeline;PortsUtil",
          "MetricName": "ILP"
      },
-    {
-        "BriefDescription": "Branch Misprediction Cost: Fraction of TMA slots wasted per non-speculative branch misprediction (retired JEClear)",
-        "MetricExpr": " ( ((BR_MISP_RETIRED.ALL_BRANCHES / ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT )) * (( UOPS_ISSUED.ANY - UOPS_RETIRED.RETIRE_SLOTS + 4 * INT_MISC.RECOVERY_CYCLES ) / (4 * CPU_CLK_UNHALTED.THREAD))) + (4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) * (BR_MISP_RETIRED.ALL_BRANCHES * (12 * ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT + BACLEARS.ANY ) / CPU_CLK_UNHALTED.THREAD) / ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT + BACLEARS.ANY )) / #(4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) ) * (4 * CPU_CLK_UNHALTED.THREAD) / BR_MISP_RETIRED.ALL_BRANCHES",
-        "MetricGroup": "Bad;BrMispredicts",
-        "MetricName": "Branch_Misprediction_Cost"
-    },
-    {
-        "BriefDescription": "Branch Misprediction Cost: Fraction of TMA slots wasted per non-speculative branch misprediction (retired JEClear)",
-        "MetricExpr": " ( ((BR_MISP_RETIRED.ALL_BRANCHES / ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT )) * (( UOPS_ISSUED.ANY - UOPS_RETIRED.RETIRE_SLOTS + 4 * ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) ) / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )))) + (4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) * (BR_MISP_RETIRED.ALL_BRANCHES * (12 * ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT + BACLEARS.ANY ) / CPU_CLK_UNHALTED.THREAD) / ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT + BACLEARS.ANY )) / #(4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) ) * (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )) / BR_MISP_RETIRED.ALL_BRANCHES",
-        "MetricGroup": "Bad;BrMispredicts_SMT",
-        "MetricName": "Branch_Misprediction_Cost_SMT"
-    },
-    {
-        "BriefDescription": "Number of Instructions per non-speculative Branch Misprediction (JEClear)",
-        "MetricExpr": "INST_RETIRED.ANY / BR_MISP_RETIRED.ALL_BRANCHES",
-        "MetricGroup": "Bad;BadSpec;BrMispredicts",
-        "MetricName": "IpMispredict"
-    },
      {
          "BriefDescription": "Core actual clocks when any Logical Processor is active on the Physical Core",
          "MetricExpr": "( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )",
@@ -256,6 +232,18 @@
          "MetricGroup": "Summary;TmaL1",
          "MetricName": "Instructions"
      },
+    {
+        "BriefDescription": "Average number of Uops retired in cycles where at least one uop has retired.",
+        "MetricExpr": "UOPS_RETIRED.RETIRE_SLOTS / cpu@UOPS_RETIRED.RETIRE_SLOTS\\,cmask\\=1@",
+        "MetricGroup": "Pipeline;Ret",
+        "MetricName": "Retire"
+    },
+    {
+        "BriefDescription": "",
+        "MetricExpr": "UOPS_EXECUTED.THREAD / cpu@UOPS_EXECUTED.THREAD\\,cmask\\=1@",
+        "MetricGroup": "Cor;Pipeline;PortsUtil;SMT",
+        "MetricName": "Execute"
+    },
      {
          "BriefDescription": "Fraction of Uops delivered by the DSB (aka Decoded ICache; or Uop Cache)",
          "MetricExpr": "IDQ.DSB_UOPS / (( IDQ.DSB_UOPS + LSD.UOPS + IDQ.MITE_UOPS + IDQ.MS_UOPS ) )",
@@ -263,35 +251,34 @@
          "MetricName": "DSB_Coverage"
      },
      {
-        "BriefDescription": "Actual Average Latency for L1 data-cache miss demand load instructions (in core cycles)",
-        "MetricExpr": "L1D_PEND_MISS.PENDING / ( MEM_LOAD_UOPS_RETIRED.L1_MISS + mem_load_uops_retired.hit_lfb )",
-        "MetricGroup": "Mem;MemoryBound;MemoryLat",
-        "MetricName": "Load_Miss_Real_Latency",
-        "PublicDescription": "Actual Average Latency for L1 data-cache miss demand load instructions (in core cycles). Latency may be overestimated for multi-load instructions - e.g. repeat strings."
+        "BriefDescription": "Number of Instructions per non-speculative Branch Misprediction (JEClear) (lower number means higher occurrence rate)",
+        "MetricExpr": "INST_RETIRED.ANY / BR_MISP_RETIRED.ALL_BRANCHES",
+        "MetricGroup": "Bad;BadSpec;BrMispredicts",
+        "MetricName": "IpMispredict"
      },
      {
-        "BriefDescription": "Memory-Level-Parallelism (average number of L1 miss demand load when there is at least one such miss. Per-Logical Processor)",
-        "MetricExpr": "L1D_PEND_MISS.PENDING / L1D_PEND_MISS.PENDING_CYCLES",
-        "MetricGroup": "Mem;MemoryBound;MemoryBW",
-        "MetricName": "MLP"
+        "BriefDescription": "Branch Misprediction Cost: Fraction of TMA slots wasted per non-speculative branch misprediction (retired JEClear)",
+        "MetricExpr": " ( ((BR_MISP_RETIRED.ALL_BRANCHES / ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT )) * (( UOPS_ISSUED.ANY - UOPS_RETIRED.RETIRE_SLOTS + 4 * INT_MISC.RECOVERY_CYCLES ) / (4 * CPU_CLK_UNHALTED.THREAD))) + (4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) * (BR_MISP_RETIRED.ALL_BRANCHES * (12 * ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT + BACLEARS.ANY ) / CPU_CLK_UNHALTED.THREAD) / ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT + BACLEARS.ANY )) / #(4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) ) * (4 * CPU_CLK_UNHALTED.THREAD) / BR_MISP_RETIRED.ALL_BRANCHES",
+        "MetricGroup": "Bad;BrMispredicts",
+        "MetricName": "Branch_Misprediction_Cost"
      },
      {
-        "BriefDescription": "Average data fill bandwidth to the L1 data cache [GB / sec]",
-        "MetricExpr": "64 * L1D.REPLACEMENT / 1000000000 / duration_time",
-        "MetricGroup": "Mem;MemoryBW",
-        "MetricName": "L1D_Cache_Fill_BW"
+        "BriefDescription": "Branch Misprediction Cost: Fraction of TMA slots wasted per non-speculative branch misprediction (retired JEClear)",
+        "MetricExpr": " ( ((BR_MISP_RETIRED.ALL_BRANCHES / ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT )) * (( UOPS_ISSUED.ANY - UOPS_RETIRED.RETIRE_SLOTS + 4 * ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) ) / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )))) + (4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) * (BR_MISP_RETIRED.ALL_BRANCHES * (12 * ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT + BACLEARS.ANY ) / CPU_CLK_UNHALTED.THREAD) / ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT + BACLEARS.ANY )) / #(4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) ) * (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )) / BR_MISP_RETIRED.ALL_BRANCHES",
+        "MetricGroup": "Bad;BrMispredicts_SMT",
+        "MetricName": "Branch_Misprediction_Cost_SMT"
      },
      {
-        "BriefDescription": "Average data fill bandwidth to the L2 cache [GB / sec]",
-        "MetricExpr": "64 * L2_LINES_IN.ALL / 1000000000 / duration_time",
-        "MetricGroup": "Mem;MemoryBW",
-        "MetricName": "L2_Cache_Fill_BW"
+        "BriefDescription": "Actual Average Latency for L1 data-cache miss demand load operations (in core cycles)",
+        "MetricExpr": "L1D_PEND_MISS.PENDING / ( MEM_LOAD_UOPS_RETIRED.L1_MISS + mem_load_uops_retired.hit_lfb )",
+        "MetricGroup": "Mem;MemoryBound;MemoryLat",
+        "MetricName": "Load_Miss_Real_Latency"
      },
      {
-        "BriefDescription": "Average per-core data fill bandwidth to the L3 cache [GB / sec]",
-        "MetricExpr": "64 * LONGEST_LAT_CACHE.MISS / 1000000000 / duration_time",
-        "MetricGroup": "Mem;MemoryBW",
-        "MetricName": "L3_Cache_Fill_BW"
+        "BriefDescription": "Memory-Level-Parallelism (average number of L1 miss demand load when there is at least one such miss. Per-Logical Processor)",
+        "MetricExpr": "L1D_PEND_MISS.PENDING / L1D_PEND_MISS.PENDING_CYCLES",
+        "MetricGroup": "Mem;MemoryBound;MemoryBW",
+        "MetricName": "MLP"
      },
      {
          "BriefDescription": "L1 cache true misses per kilo instruction for retired demand loads",
@@ -306,13 +293,13 @@
          "MetricName": "L2MPKI"
      },
      {
-        "BriefDescription": "L2 cache misses per kilo instruction for all request types (including speculative)",
+        "BriefDescription": "L2 cache ([RKL+] true) misses per kilo instruction for all request types (including speculative)",
          "MetricExpr": "1000 * L2_RQSTS.MISS / INST_RETIRED.ANY",
          "MetricGroup": "Mem;CacheMisses;Offcore",
          "MetricName": "L2MPKI_All"
      },
      {
-        "BriefDescription": "L2 cache misses per kilo instruction for all demand loads  (including speculative)",
+        "BriefDescription": "L2 cache ([RKL+] true) misses per kilo instruction for all demand loads  (including speculative)",
          "MetricExpr": "1000 * L2_RQSTS.DEMAND_DATA_RD_MISS / INST_RETIRED.ANY",
          "MetricGroup": "Mem;CacheMisses",
          "MetricName": "L2MPKI_Load"
@@ -348,6 +335,48 @@
          "MetricGroup": "Mem;MemoryTLB_SMT",
          "MetricName": "Page_Walks_Utilization_SMT"
      },
+    {
+        "BriefDescription": "Average per-core data fill bandwidth to the L1 data cache [GB / sec]",
+        "MetricExpr": "64 * L1D.REPLACEMENT / 1000000000 / duration_time",
+        "MetricGroup": "Mem;MemoryBW",
+        "MetricName": "L1D_Cache_Fill_BW"
+    },
+    {
+        "BriefDescription": "Average per-core data fill bandwidth to the L2 cache [GB / sec]",
+        "MetricExpr": "64 * L2_LINES_IN.ALL / 1000000000 / duration_time",
+        "MetricGroup": "Mem;MemoryBW",
+        "MetricName": "L2_Cache_Fill_BW"
+    },
+    {
+        "BriefDescription": "Average per-core data fill bandwidth to the L3 cache [GB / sec]",
+        "MetricExpr": "64 * LONGEST_LAT_CACHE.MISS / 1000000000 / duration_time",
+        "MetricGroup": "Mem;MemoryBW",
+        "MetricName": "L3_Cache_Fill_BW"
+    },
+    {
+        "BriefDescription": "Average per-thread data fill bandwidth to the L1 data cache [GB / sec]",
+        "MetricExpr": "(64 * L1D.REPLACEMENT / 1000000000 / duration_time)",
+        "MetricGroup": "Mem;MemoryBW",
+        "MetricName": "L1D_Cache_Fill_BW_1T"
+    },
+    {
+        "BriefDescription": "Average per-thread data fill bandwidth to the L2 cache [GB / sec]",
+        "MetricExpr": "(64 * L2_LINES_IN.ALL / 1000000000 / duration_time)",
+        "MetricGroup": "Mem;MemoryBW",
+        "MetricName": "L2_Cache_Fill_BW_1T"
+    },
+    {
+        "BriefDescription": "Average per-thread data fill bandwidth to the L3 cache [GB / sec]",
+        "MetricExpr": "(64 * LONGEST_LAT_CACHE.MISS / 1000000000 / duration_time)",
+        "MetricGroup": "Mem;MemoryBW",
+        "MetricName": "L3_Cache_Fill_BW_1T"
+    },
+    {
+        "BriefDescription": "Average per-thread data access bandwidth to the L3 cache [GB / sec]",
+        "MetricExpr": "0",
+        "MetricGroup": "Mem;MemoryBW;Offcore",
+        "MetricName": "L3_Cache_Access_BW_1T"
+    },
      {
          "BriefDescription": "Average CPU Utilization",
          "MetricExpr": "CPU_CLK_UNHALTED.REF_TSC / msr@tsc@",
@@ -364,7 +393,8 @@
          "BriefDescription": "Giga Floating Point Operations Per Second",
          "MetricExpr": "( ( 1 * ( FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE ) + 2 * FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + 4 * ( FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE ) + 8 * FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE ) / 1000000000 ) / duration_time",
          "MetricGroup": "Cor;Flops;HPC",
-        "MetricName": "GFLOPs"
+        "MetricName": "GFLOPs",
+        "PublicDescription": "Giga Floating Point Operations Per Second. Aggregate across all supported options of: FP precisions, scalar and vector instructions, vector-width and AMX engine."
      },
      {
          "BriefDescription": "Average Frequency Utilization relative nominal frequency",
@@ -461,5 +491,439 @@
          "MetricExpr": "(cstate_pkg@c7\\-residency@ / msr@tsc@) * 100",
          "MetricGroup": "Power",
          "MetricName": "C7_Pkg_Residency"
+    },
+    {
+        "BriefDescription": "CPU operating frequency (in GHz)",
+        "MetricExpr": "( CPU_CLK_UNHALTED.THREAD / CPU_CLK_UNHALTED.REF_TSC * #SYSTEM_TSC_FREQ ) / 1000000000",
+        "MetricGroup": "",
+        "MetricName": "cpu_operating_frequency",
+        "ScaleUnit": "1GHz"
+    },
+    {
+        "BriefDescription": "Cycles per instruction retired; indicating how much time each executed instruction took; in units of cycles.",
+        "MetricExpr": "CPU_CLK_UNHALTED.THREAD / INST_RETIRED.ANY",
+        "MetricGroup": "",
+        "MetricName": "cpi",
+        "ScaleUnit": "1per_instr"
+    },
+    {
+        "BriefDescription": "The ratio of number of completed memory load instructions to the total number completed instructions",
+        "MetricExpr": "MEM_UOPS_RETIRED.ALL_LOADS / INST_RETIRED.ANY",
+        "MetricGroup": "",
+        "MetricName": "loads_per_instr",
+        "ScaleUnit": "1per_instr"
+    },
+    {
+        "BriefDescription": "The ratio of number of completed memory store instructions to the total number completed instructions",
+        "MetricExpr": "MEM_UOPS_RETIRED.ALL_STORES / INST_RETIRED.ANY",
+        "MetricGroup": "",
+        "MetricName": "stores_per_instr",
+        "ScaleUnit": "1per_instr"
+    },
+    {
+        "BriefDescription": "Ratio of number of requests missing L1 data cache (includes data+rfo w/ prefetches) to the total number of completed instructions",
+        "MetricExpr": "L1D.REPLACEMENT / INST_RETIRED.ANY",
+        "MetricGroup": "",
+        "MetricName": "l1d_mpi_includes_data_plus_rfo_with_prefetches",
+        "ScaleUnit": "1per_instr"
+    },
+    {
+        "BriefDescription": "Ratio of number of demand load requests hitting in L1 data cache to the total number of completed instructions",
+        "MetricExpr": "MEM_LOAD_UOPS_RETIRED.L1_HIT / INST_RETIRED.ANY",
+        "MetricGroup": "",
+        "MetricName": "l1d_demand_data_read_hits_per_instr",
+        "ScaleUnit": "1per_instr"
+    },
+    {
+        "BriefDescription": "Ratio of number of code read requests missing in L1 instruction cache (includes prefetches) to the total number of completed instructions",
+        "MetricExpr": "L2_RQSTS.ALL_CODE_RD / INST_RETIRED.ANY",
+        "MetricGroup": "",
+        "MetricName": "l1_i_code_read_misses_with_prefetches_per_instr",
+        "ScaleUnit": "1per_instr"
+    },
+    {
+        "BriefDescription": "Ratio of number of completed demand load requests hitting in L2 cache to the total number of completed instructions",
+        "MetricExpr": "MEM_LOAD_UOPS_RETIRED.L2_HIT / INST_RETIRED.ANY",
+        "MetricGroup": "",
+        "MetricName": "l2_demand_data_read_hits_per_instr",
+        "ScaleUnit": "1per_instr"
+    },
+    {
+        "BriefDescription": "Ratio of number of requests missing L2 cache (includes code+data+rfo w/ prefetches) to the total number of completed instructions",
+        "MetricExpr": "L2_LINES_IN.ALL / INST_RETIRED.ANY",
+        "MetricGroup": "",
+        "MetricName": "l2_mpi_includes_code_plus_data_plus_rfo_with_prefetches",
+        "ScaleUnit": "1per_instr"
+    },
+    {
+        "BriefDescription": "Ratio of number of completed data read request missing L2 cache to the total number of completed instructions",
+        "MetricExpr": "MEM_LOAD_UOPS_RETIRED.L2_MISS / INST_RETIRED.ANY",
+        "MetricGroup": "",
+        "MetricName": "l2_demand_data_read_mpi",
+        "ScaleUnit": "1per_instr"
+    },
+    {
+        "BriefDescription": "Ratio of number of code read request missing L2 cache to the total number of completed instructions",
+        "MetricExpr": "L2_RQSTS.CODE_RD_MISS / INST_RETIRED.ANY",
+        "MetricGroup": "",
+        "MetricName": "l2_demand_code_mpi",
+        "ScaleUnit": "1per_instr"
+    },
+    {
+        "BriefDescription": "Ratio of number of data read requests missing last level core cache (includes demand w/ prefetches) to the total number of completed instructions",
+        "MetricExpr": "( cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x182@ + cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x192@ ) / INST_RETIRED.ANY",
+        "MetricGroup": "",
+        "MetricName": "llc_data_read_mpi_demand_plus_prefetch",
+        "ScaleUnit": "1per_instr"
+    },
+    {
+        "BriefDescription": "Ratio of number of code read requests missing last level core cache (includes demand w/ prefetches) to the total number of completed instructions",
+        "MetricExpr": "( cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x181@ + cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x191@ ) / INST_RETIRED.ANY",
+        "MetricGroup": "",
+        "MetricName": "llc_code_read_mpi_demand_plus_prefetch",
+        "ScaleUnit": "1per_instr"
+    },
+    {
+        "BriefDescription": "Average latency of a last level cache (LLC) demand and prefetch data read miss (read memory access) in nano seconds",
+        "MetricExpr": "( 1000000000 * ( cbox@UNC_C_TOR_OCCUPANCY.MISS_OPCODE\\,filter_opc\\=0x182@ / cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x182@ ) / ( UNC_C_CLOCKTICKS / ( source_count(UNC_C_CLOCKTICKS) * #num_packages ) ) ) * duration_time",
+        "MetricGroup": "",
+        "MetricName": "llc_data_read_demand_plus_prefetch_miss_latency",
+        "ScaleUnit": "1ns"
+    },
+    {
+        "BriefDescription": "Average latency of a last level cache (LLC) demand and prefetch data read miss (read memory access) addressed to local memory in nano seconds",
+        "MetricExpr": "( 1000000000 * ( cbox@UNC_C_TOR_OCCUPANCY.MISS_LOCAL_OPCODE\\,filter_opc\\=0x182@ / cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x182@ ) / ( UNC_C_CLOCKTICKS / ( source_count(UNC_C_CLOCKTICKS) * #num_packages ) ) ) * duration_time",
+        "MetricGroup": "",
+        "MetricName": "llc_data_read_demand_plus_prefetch_miss_latency_for_local_requests",
+        "ScaleUnit": "1ns"
+    },
+    {
+        "BriefDescription": "Average latency of a last level cache (LLC) demand and prefetch data read miss (read memory access) addressed to remote memory in nano seconds",
+        "MetricExpr": "( 1000000000 * ( cbox@UNC_C_TOR_OCCUPANCY.MISS_REMOTE_OPCODE\\,filter_opc\\=0x182@ / cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x182@ ) / ( UNC_C_CLOCKTICKS / ( source_count(UNC_C_CLOCKTICKS) * #num_packages ) ) ) * duration_time",
+        "MetricGroup": "",
+        "MetricName": "llc_data_read_demand_plus_prefetch_miss_latency_for_remote_requests",
+        "ScaleUnit": "1ns"
+    },
+    {
+        "BriefDescription": "Ratio of number of completed page walks (for all page sizes) caused by a code fetch to the total number of completed instructions. This implies it missed in the ITLB (Instruction TLB) and further levels of TLB.",
+        "MetricExpr": "ITLB_MISSES.WALK_COMPLETED / INST_RETIRED.ANY",
+        "MetricGroup": "",
+        "MetricName": "itlb_mpi",
+        "ScaleUnit": "1per_instr"
+    },
+    {
+        "BriefDescription": "Ratio of number of completed page walks (for 2 megabyte and 4 megabyte page sizes) caused by a code fetch to the total number of completed instructions. This implies it missed in the Instruction Translation Lookaside Buffer (ITLB) and further levels of TLB.",
+        "MetricExpr": "ITLB_MISSES.WALK_COMPLETED_2M_4M / INST_RETIRED.ANY",
+        "MetricGroup": "",
+        "MetricName": "itlb_large_page_mpi",
+        "ScaleUnit": "1per_instr"
+    },
+    {
+        "BriefDescription": "Ratio of number of completed page walks (for all page sizes) caused by demand data loads to the total number of completed instructions. This implies it missed in the DTLB and further levels of TLB.",
+        "MetricExpr": "DTLB_LOAD_MISSES.WALK_COMPLETED / INST_RETIRED.ANY",
+        "MetricGroup": "",
+        "MetricName": "dtlb_load_mpi",
+        "ScaleUnit": "1per_instr"
+    },
+    {
+        "BriefDescription": "Ratio of number of completed page walks (for all page sizes) caused by demand data stores to the total number of completed instructions. This implies it missed in the DTLB and further levels of TLB.",
+        "MetricExpr": "DTLB_STORE_MISSES.WALK_COMPLETED / INST_RETIRED.ANY",
+        "MetricGroup": "",
+        "MetricName": "dtlb_store_mpi",
+        "ScaleUnit": "1per_instr"
+    },
+    {
+        "BriefDescription": "Memory read that miss the last level cache (LLC) addressed to local DRAM as a percentage of total memory read accesses, does not include LLC prefetches.",
+        "MetricExpr": "100 * cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x182@ / ( cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x182@ + cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x182@ )",
+        "MetricGroup": "",
+        "MetricName": "numa_percent_reads_addressed_to_local_dram",
+        "ScaleUnit": "1%"
+    },
+    {
+        "BriefDescription": "Memory reads that miss the last level cache (LLC) addressed to remote DRAM as a percentage of total memory read accesses, does not include LLC prefetches.",
+        "MetricExpr": "100 * cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x182@ / ( cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x182@ + cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x182@ )",
+        "MetricGroup": "",
+        "MetricName": "numa_percent_reads_addressed_to_remote_dram",
+        "ScaleUnit": "1%"
+    },
+    {
+        "BriefDescription": "Uncore operating frequency in GHz",
+        "MetricExpr": "UNC_C_CLOCKTICKS / ( source_count(UNC_C_CLOCKTICKS) * #num_packages ) / 1000000000",
+        "MetricGroup": "",
+        "MetricName": "uncore_frequency",
+        "ScaleUnit": "1GHz"
+    },
+    {
+        "BriefDescription": "Intel(R) Quick Path Interconnect (QPI) data transmit bandwidth (MB/sec)",
+        "MetricExpr": "( UNC_Q_TxL_FLITS_G0.DATA * 8 / 1000000) / duration_time",
+        "MetricGroup": "",
+        "MetricName": "qpi_data_transmit_bw_only_data",
+        "ScaleUnit": "1MB/s"
+    },
+    {
+        "BriefDescription": "DDR memory read bandwidth (MB/sec)",
+        "MetricExpr": "( UNC_M_CAS_COUNT.RD * 64 / 1000000) / duration_time",
+        "MetricGroup": "",
+        "MetricName": "memory_bandwidth_read",
+        "ScaleUnit": "1MB/s"
+    },
+    {
+        "BriefDescription": "DDR memory write bandwidth (MB/sec)",
+        "MetricExpr": "( UNC_M_CAS_COUNT.WR * 64 / 1000000) / duration_time",
+        "MetricGroup": "",
+        "MetricName": "memory_bandwidth_write",
+        "ScaleUnit": "1MB/s"
+    },
+    {
+        "BriefDescription": "DDR memory bandwidth (MB/sec)",
+        "MetricExpr": "(( UNC_M_CAS_COUNT.RD + UNC_M_CAS_COUNT.WR ) * 64 / 1000000) / duration_time",
+        "MetricGroup": "",
+        "MetricName": "memory_bandwidth_total",
+        "ScaleUnit": "1MB/s"
+    },
+    {
+        "BriefDescription": "Bandwidth of IO reads that are initiated by end device controllers that are requesting memory from the CPU.",
+        "MetricExpr": "( cbox@UNC_C_TOR_INSERTS.OPCODE\\,filter_opc\\=0x19e@ * 64 / 1000000) / duration_time",
+        "MetricGroup": "",
+        "MetricName": "io_bandwidth_read",
+        "ScaleUnit": "1MB/s"
+    },
+    {
+        "BriefDescription": "Bandwidth of IO writes that are initiated by end device controllers that are writing memory to the CPU.",
+        "MetricExpr": "(( cbox@UNC_C_TOR_INSERTS.OPCODE\\,filter_opc\\=0x1c8\\,filter_tid\\=0x3e@ + cbox@UNC_C_TOR_INSERTS.OPCODE\\,filter_opc\\=0x180\\,filter_tid\\=0x3e@ ) * 64 / 1000000) / duration_time",
+        "MetricGroup": "",
+        "MetricName": "io_bandwidth_write",
+        "ScaleUnit": "1MB/s"
+    },
+    {
+        "BriefDescription": "Uops delivered from decoded instruction cache (decoded stream buffer or DSB) as a percent of total uops delivered to Instruction Decode Queue",
+        "MetricExpr": "100 * ( IDQ.DSB_UOPS / UOPS_ISSUED.ANY )",
+        "MetricGroup": "",
+        "MetricName": "percent_uops_delivered_from_decoded_icache_dsb",
+        "ScaleUnit": "1%"
+    },
+    {
+        "BriefDescription": "Uops delivered from legacy decode pipeline (Micro-instruction Translation Engine or MITE) as a percent of total uops delivered to Instruction Decode Queue",
+        "MetricExpr": "100 * ( IDQ.MITE_UOPS / UOPS_ISSUED.ANY )",
+        "MetricGroup": "",
+        "MetricName": "percent_uops_delivered_from_legacy_decode_pipeline_mite",
+        "ScaleUnit": "1%"
+    },
+    {
+        "BriefDescription": "Uops delivered from microcode sequencer (MS) as a percent of total uops delivered to Instruction Decode Queue",
+        "MetricExpr": "100 * ( IDQ.MS_UOPS / UOPS_ISSUED.ANY )",
+        "MetricGroup": "",
+        "MetricName": "percent_uops_delivered_from_microcode_sequencer_ms",
+        "ScaleUnit": "1%"
+    },
+    {
+        "BriefDescription": "Uops delivered from loop stream detector(LSD) as a percent of total uops delivered to Instruction Decode Queue",
+        "MetricExpr": "100 * ( LSD.UOPS / UOPS_ISSUED.ANY )",
+        "MetricGroup": "",
+        "MetricName": "percent_uops_delivered_from_loop_stream_detector_lsd",
+        "ScaleUnit": "1%"
+    },
+    {
+        "BriefDescription": "This category represents fraction of slots where the processor's Frontend undersupplies its Backend. Frontend denotes the first part of the processor core responsible to fetch operations that are executed later on by the Backend part. Within the Frontend; a branch predictor predicts the next address to fetch; cache-lines are fetched from the memory subsystem; parsed into instructions; and lastly decoded into micro-operations (uops). Ideally the Frontend can issue Machine_Width uops every cycle to the Backend. Frontend Bound denotes unutilized issue-slots when there is no Backend stall; i.e. bubbles where Frontend delivered no uops while Backend could have accepted them. For example; stalls due to instruction-cache misses would be categorized under Frontend Bound.",
+        "MetricExpr": "100 * ( IDQ_UOPS_NOT_DELIVERED.CORE / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) )",
+        "MetricGroup": "TmaL1;PGO",
+        "MetricName": "tma_frontend_bound_percent",
+        "ScaleUnit": "1%"
+    },
+    {
+        "BriefDescription": "This metric represents fraction of slots the CPU was stalled due to Frontend latency issues.  For example; instruction-cache misses; iTLB misses or fetch stalls after a branch misprediction are categorized under Frontend Latency. In such cases; the Frontend eventually delivers no uops for some period.",
+        "MetricExpr": "100 * ( ( 4 ) * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) )",
+        "MetricGroup": "Frontend;TmaL2;m_tma_frontend_bound_percent",
+        "MetricName": "tma_fetch_latency_percent",
+        "ScaleUnit": "1%"
+    },
+    {
+        "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to instruction cache misses.",
+        "MetricExpr": "100 * ( ICACHE.IFDATA_STALL / ( CPU_CLK_UNHALTED.THREAD ) )",
+        "MetricGroup": "BigFoot;FetchLat;IcMiss;TmaL3;m_tma_fetch_latency_percent",
+        "MetricName": "tma_icache_misses_percent",
+        "ScaleUnit": "1%"
+    },
+    {
+        "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to Instruction TLB (ITLB) misses.",
+        "MetricExpr": "100 * ( ( 14 * ITLB_MISSES.STLB_HIT + cpu@ITLB_MISSES.WALK_DURATION\\,cmask\\=0x1@ + 7 * ITLB_MISSES.WALK_COMPLETED ) / ( CPU_CLK_UNHALTED.THREAD ) )",
+        "MetricGroup": "BigFoot;FetchLat;MemoryTLB;TmaL3;m_tma_fetch_latency_percent",
+        "MetricName": "tma_itlb_misses_percent",
+        "ScaleUnit": "1%"
+    },
+    {
+        "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to Branch Resteers. Branch Resteers estimates the Frontend delay in fetching operations from corrected path; following all sorts of miss-predicted branches. For example; branchy code with lots of miss-predictions might get categorized under Branch Resteers. Note the value of this node may overlap with its siblings.",
+        "MetricExpr": "100 * ( ( 12 ) * ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT + BACLEARS.ANY ) / ( CPU_CLK_UNHALTED.THREAD ) )",
+        "MetricGroup": "FetchLat;TmaL3;m_tma_fetch_latency_percent",
+        "MetricName": "tma_branch_resteers_percent",
+        "ScaleUnit": "1%"
+    },
+    {
+        "BriefDescription": "This metric represents fraction of cycles the CPU was stalled due to switches from DSB to MITE pipelines. The DSB (decoded i-cache) is a Uop Cache where the front-end directly delivers Uops (micro operations) avoiding heavy x86 decoding. The DSB pipeline has shorter latency and delivered higher bandwidth than the MITE (legacy instruction decode pipeline). Switching between the two pipelines can cause penalties hence this metric measures the exposed penalty.",
+        "MetricExpr": "100 * ( DSB2MITE_SWITCHES.PENALTY_CYCLES / ( CPU_CLK_UNHALTED.THREAD ) )",
+        "MetricGroup": "DSBmiss;FetchLat;TmaL3;m_tma_fetch_latency_percent",
+        "MetricName": "tma_dsb_switches_percent",
+        "ScaleUnit": "1%"
+    },
+    {
+        "BriefDescription": "This metric represents fraction of cycles CPU was stalled due to Length Changing Prefixes (LCPs). Using proper compiler flags or Intel Compiler by default will certainly avoid this. #Link: Optimization Guide about LCP BKMs.",
+        "MetricExpr": "100 * ( ILD_STALL.LCP / ( CPU_CLK_UNHALTED.THREAD ) )",
+        "MetricGroup": "FetchLat;TmaL3;m_tma_fetch_latency_percent",
+        "MetricName": "tma_lcp_percent",
+        "ScaleUnit": "1%"
+    },
+    {
+        "BriefDescription": "This metric estimates the fraction of cycles when the CPU was stalled due to switches of uop delivery to the Microcode Sequencer (MS). Commonly used instructions are optimized for delivery by the DSB (decoded i-cache) or MITE (legacy instruction decode) pipelines. Certain operations cannot be handled natively by the execution pipeline; and must be performed by microcode (small programs injected into the execution stream). Switching to the MS too often can negatively impact performance. The MS is designated to deliver long uop flows required by CISC instructions like CPUID; or uncommon conditions like Floating Point Assists when dealing with Denormals.",
+        "MetricExpr": "100 * ( ( 2 ) * IDQ.MS_SWITCHES / ( CPU_CLK_UNHALTED.THREAD ) )",
+        "MetricGroup": "FetchLat;MicroSeq;TmaL3;m_tma_fetch_latency_percent",
+        "MetricName": "tma_ms_switches_percent",
+        "ScaleUnit": "1%"
+    },
+    {
+        "BriefDescription": "This metric represents fraction of slots the CPU was stalled due to Frontend bandwidth issues.  For example; inefficiencies at the instruction decoders; or restrictions for caching in the DSB (decoded uops cache) are categorized under Fetch Bandwidth. In such cases; the Frontend typically delivers suboptimal amount of uops to the Backend.",
+        "MetricExpr": "100 * ( ( IDQ_UOPS_NOT_DELIVERED.CORE / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) - ( ( 4 ) * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) )",
+        "MetricGroup": "FetchBW;Frontend;TmaL2;m_tma_frontend_bound_percent",
+        "MetricName": "tma_fetch_bandwidth_percent",
+        "ScaleUnit": "1%"
+    },
+    {
+        "BriefDescription": "This metric represents Core fraction of cycles in which CPU was likely limited due to the MITE pipeline (the legacy decode pipeline). This pipeline is used for code that was not pre-cached in the DSB or LSD. For example; inefficiencies due to asymmetric decoders; use of long immediate or LCP can manifest as MITE fetch bandwidth bottleneck.",
+        "MetricExpr": "100 * ( ( IDQ.ALL_MITE_CYCLES_ANY_UOPS - IDQ.ALL_MITE_CYCLES_4_UOPS ) / ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) / 2 )",
+        "MetricGroup": "DSBmiss;FetchBW;TmaL3;m_tma_fetch_bandwidth_percent",
+        "MetricName": "tma_mite_percent",
+        "ScaleUnit": "1%"
+    },
+    {
+        "BriefDescription": "This metric represents Core fraction of cycles in which CPU was likely limited due to DSB (decoded uop cache) fetch pipeline.  For example; inefficient utilization of the DSB cache structure or bank conflict when reading from it; are categorized here.",
+        "MetricExpr": "100 * ( ( IDQ.ALL_DSB_CYCLES_ANY_UOPS - IDQ.ALL_DSB_CYCLES_4_UOPS ) / ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) / 2 )",
+        "MetricGroup": "DSB;FetchBW;TmaL3;m_tma_fetch_bandwidth_percent",
+        "MetricName": "tma_dsb_percent",
+        "ScaleUnit": "1%"
+    },
+    {
+        "BriefDescription": "This category represents fraction of slots wasted due to incorrect speculations. This include slots used to issue uops that do not eventually get retired and slots for which the issue-pipeline was blocked due to recovery from earlier incorrect speculation. For example; wasted work due to miss-predicted branches are categorized under Bad Speculation category. Incorrect data speculation followed by Memory Ordering Nukes is another example.",
+        "MetricExpr": "100 * ( ( UOPS_ISSUED.ANY - ( UOPS_RETIRED.RETIRE_SLOTS ) + ( 4 ) * ( ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) if #SMT_on else INT_MISC.RECOVERY_CYCLES ) ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) )",
+        "MetricGroup": "TmaL1",
+        "MetricName": "tma_bad_speculation_percent",
+        "ScaleUnit": "1%"
+    },
+    {
+        "BriefDescription": "This metric represents fraction of slots the CPU has wasted due to Branch Misprediction.  These slots are either wasted by uops fetched from an incorrectly speculated program path; or stalls when the out-of-order part of the machine needs to recover its state from a speculative path.",
+        "MetricExpr": "100 * ( ( BR_MISP_RETIRED.ALL_BRANCHES / ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT ) ) * ( ( UOPS_ISSUED.ANY - ( UOPS_RETIRED.RETIRE_SLOTS ) + ( 4 ) * ( ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) if #SMT_on else INT_MISC.RECOVERY_CYCLES ) ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) )",
+        "MetricGroup": "BadSpec;BrMispredicts;TmaL2;m_tma_bad_speculation_percent",
+        "MetricName": "tma_branch_mispredicts_percent",
+        "ScaleUnit": "1%"
+    },
+    {
+        "BriefDescription": "This metric represents fraction of slots the CPU has wasted due to Machine Clears.  These slots are either wasted by uops fetched prior to the clear; or stalls the out-of-order portion of the machine needs to recover its state after the clear. For example; this can happen due to memory ordering Nukes (e.g. Memory Disambiguation) or Self-Modifying-Code (SMC) nukes.",
+        "MetricExpr": "100 * ( ( ( UOPS_ISSUED.ANY - ( UOPS_RETIRED.RETIRE_SLOTS ) + ( 4 ) * ( ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) if #SMT_on else INT_MISC.RECOVERY_CYCLES ) ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) - ( ( BR_MISP_RETIRED.ALL_BRANCHES / ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT ) ) * ( ( UOPS_ISSUED.ANY - ( UOPS_RETIRED.RETIRE_SLOTS ) + ( 4 ) * ( ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) if #SMT_on else INT_MISC.RECOVERY_CYCLES ) ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) ) )",
+        "MetricGroup": "BadSpec;MachineClears;TmaL2;m_tma_bad_speculation_percent",
+        "MetricName": "tma_machine_clears_percent",
+        "ScaleUnit": "1%"
+    },
+    {
+        "BriefDescription": "This category represents fraction of slots where no uops are being delivered due to a lack of required resources for accepting new uops in the Backend. Backend is the portion of the processor core where the out-of-order scheduler dispatches ready uops into their respective execution units; and once completed these uops get retired according to program order. For example; stalls due to data-cache misses or stalls due to the divider unit being overloaded are both categorized under Backend Bound. Backend Bound is further divided into two main categories: Memory Bound and Core Bound.",
+        "MetricExpr": "100 * ( 1 - ( ( IDQ_UOPS_NOT_DELIVERED.CORE / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) + ( ( UOPS_ISSUED.ANY - ( UOPS_RETIRED.RETIRE_SLOTS ) + ( 4 ) * ( ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) if #SMT_on else INT_MISC.RECOVERY_CYCLES ) ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) + ( ( UOPS_RETIRED.RETIRE_SLOTS ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) ) )",
+        "MetricGroup": "TmaL1",
+        "MetricName": "tma_backend_bound_percent",
+        "ScaleUnit": "1%"
+    },
+    {
+        "BriefDescription": "This metric represents fraction of slots the Memory subsystem within the Backend was a bottleneck.  Memory Bound estimates fraction of slots where pipeline is likely stalled due to demand load or store instructions. This accounts mainly for (1) non-completed in-flight memory demand loads which coincides with execution units starvation; in addition to (2) cases where stores could impose backpressure on the pipeline when many of them get buffered at the same time (less common out of the two).",
+        "MetricExpr": "100 * ( ( ( CYCLE_ACTIVITY.STALLS_MEM_ANY + RESOURCE_STALLS.SB ) / ( ( CYCLE_ACTIVITY.STALLS_TOTAL + UOPS_EXECUTED.CYCLES_GE_1_UOP_EXEC - ( UOPS_EXECUTED.CYCLES_GE_3_UOPS_EXEC if ( ( INST_RETIRED.ANY / ( CPU_CLK_UNHALTED.THREAD ) ) > 1.8 ) else UOPS_EXECUTED.CYCLES_GE_2_UOPS_EXEC ) - ( RS_EVENTS.EMPTY_CYCLES if ( ( ( 4 ) * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) > 0.1 ) else 0 ) + RESOURCE_STALLS.SB ) ) ) * ( 1 - ( ( IDQ_UOPS_NOT_DELIVERED.CORE / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) + ( ( UOPS_ISSUED.ANY - ( UOPS_RETIRED.RETIRE_SLOTS ) + ( 4 ) * ( ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) if #SMT_on else INT_MISC.RECOVERY_CYCLES ) ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) + ( ( UOPS_RETIRED.RETIRE_SLOTS ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) ) ) )",
+        "MetricGroup": "Backend;TmaL2;m_tma_backend_bound_percent",
+        "MetricName": "tma_memory_bound_percent",
+        "ScaleUnit": "1%"
+    },
+    {
+        "BriefDescription": "This metric estimates how often the CPU was stalled without loads missing the L1 data cache.  The L1 data cache typically has the shortest latency.  However; in certain cases like loads blocked on older stores; a load might suffer due to high latency even though it is being satisfied by the L1. Another example is loads who miss in the TLB. These cases are characterized by execution unit stalls; while some non-completed demand load lives in the machine without having that demand load missing the L1 cache.",
+        "MetricExpr": "100 * ( max( ( CYCLE_ACTIVITY.STALLS_MEM_ANY - CYCLE_ACTIVITY.STALLS_L1D_MISS ) / ( CPU_CLK_UNHALTED.THREAD ) , 0 ) )",
+        "MetricGroup": "CacheMisses;MemoryBound;TmaL3mem;TmaL3;m_tma_memory_bound_percent",
+        "MetricName": "tma_l1_bound_percent",
+        "ScaleUnit": "1%"
+    },
+    {
+        "BriefDescription": "This metric estimates how often the CPU was stalled due to L2 cache accesses by loads.  Avoiding cache misses (i.e. L1 misses/L2 hits) can improve the latency and increase performance.",
+        "MetricExpr": "100 * ( ( CYCLE_ACTIVITY.STALLS_L1D_MISS - CYCLE_ACTIVITY.STALLS_L2_MISS ) / ( CPU_CLK_UNHALTED.THREAD ) )",
+        "MetricGroup": "CacheMisses;MemoryBound;TmaL3mem;TmaL3;m_tma_memory_bound_percent",
+        "MetricName": "tma_l2_bound_percent",
+        "ScaleUnit": "1%"
+    },
+    {
+        "BriefDescription": "This metric estimates how often the CPU was stalled due to loads accesses to L3 cache or contended with a sibling Core.  Avoiding cache misses (i.e. L2 misses/L3 hits) can improve the latency and increase performance.",
+        "MetricExpr": "100 * ( ( MEM_LOAD_UOPS_RETIRED.L3_HIT / ( MEM_LOAD_UOPS_RETIRED.L3_HIT + ( 7 ) * MEM_LOAD_UOPS_RETIRED.L3_MISS ) ) * CYCLE_ACTIVITY.STALLS_L2_MISS / ( CPU_CLK_UNHALTED.THREAD ) )",
+        "MetricGroup": "CacheMisses;MemoryBound;TmaL3mem;TmaL3;m_tma_memory_bound_percent",
+        "MetricName": "tma_l3_bound_percent",
+        "ScaleUnit": "1%"
+    },
+    {
+        "BriefDescription": "This metric estimates how often the CPU was stalled on accesses to external memory (DRAM) by loads. Better caching can improve the latency and increase performance.",
+        "MetricExpr": "100 * ( min( ( ( 1 - ( MEM_LOAD_UOPS_RETIRED.L3_HIT / ( MEM_LOAD_UOPS_RETIRED.L3_HIT + ( 7 ) * MEM_LOAD_UOPS_RETIRED.L3_MISS ) ) ) * CYCLE_ACTIVITY.STALLS_L2_MISS / ( CPU_CLK_UNHALTED.THREAD ) ) , ( 1 ) ) )",
+        "MetricGroup": "MemoryBound;TmaL3mem;TmaL3;m_tma_memory_bound_percent",
+        "MetricName": "tma_dram_bound_percent",
+        "ScaleUnit": "1%"
+    },
+    {
+        "BriefDescription": "This metric estimates how often CPU was stalled  due to RFO store memory accesses; RFO store issue a read-for-ownership request before the write. Even though store accesses do not typically stall out-of-order CPUs; there are few cases where stores can lead to actual stalls. This metric will be flagged should RFO stores be a bottleneck.",
+        "MetricExpr": "100 * ( RESOURCE_STALLS.SB / ( CPU_CLK_UNHALTED.THREAD ) )",
+        "MetricGroup": "MemoryBound;TmaL3mem;TmaL3;m_tma_memory_bound_percent",
+        "MetricName": "tma_store_bound_percent",
+        "ScaleUnit": "1%"
+    },
+    {
+        "BriefDescription": "This metric represents fraction of slots where Core non-memory issues were of a bottleneck.  Shortage in hardware compute resources; or dependencies in software's instructions are both categorized under Core Bound. Hence it may indicate the machine ran out of an out-of-order resource; certain execution units are overloaded or dependencies in program's data- or instruction-flow are limiting the performance (e.g. FP-chained long-latency arithmetic operations).",
+        "MetricExpr": "100 * ( ( 1 - ( ( IDQ_UOPS_NOT_DELIVERED.CORE / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) + ( ( UOPS_ISSUED.ANY - ( UOPS_RETIRED.RETIRE_SLOTS ) + ( 4 ) * ( ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) if #SMT_on else INT_MISC.RECOVERY_CYCLES ) ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) + ( ( UOPS_RETIRED.RETIRE_SLOTS ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) ) ) - ( ( ( CYCLE_ACTIVITY.STALLS_MEM_ANY + RESOURCE_STALLS.SB ) / ( ( CYCLE_ACTIVITY.STALLS_TOTAL + UOPS_EXECUTED.CYCLES_GE_1_UOP_EXEC - ( UOPS_EXECUTED.CYCLES_GE_3_UOPS_EXEC if ( ( INST_RETIRED.ANY / ( CPU_CLK_UNHALTED.THREAD ) ) > 1.8 ) else UOPS_EXECUTED.CYCLES_GE_2_UOPS_EXEC ) - ( RS_EVENTS.EMPTY_CYCLES if ( ( ( 4 ) * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) > 0.1 ) else 0 ) + RESOURCE_STALLS.SB ) ) ) * ( 1 - ( ( IDQ_UOPS_NOT_DELIVERED.CORE / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) + ( ( UOPS_ISSUED.ANY - ( UOPS_RETIRED.RETIRE_SLOTS ) + ( 4 ) * ( ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) if #SMT_on else INT_MISC.RECOVERY_CYCLES ) ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) + ( ( UOPS_RETIRED.RETIRE_SLOTS ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) ) ) ) )",
+        "MetricGroup": "Backend;TmaL2;Compute;m_tma_backend_bound_percent",
+        "MetricName": "tma_core_bound_percent",
+        "ScaleUnit": "1%"
+    },
+    {
+        "BriefDescription": "This metric represents fraction of cycles where the Divider unit was active. Divide and square root instructions are performed by the Divider unit and can take considerably longer latency than integer or Floating Point addition; subtraction; or multiplication.",
+        "MetricExpr": "100 * ( ARITH.FPU_DIV_ACTIVE / ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) )",
+        "MetricGroup": "TmaL3;m_tma_core_bound_percent",
+        "MetricName": "tma_divider_percent",
+        "ScaleUnit": "1%"
+    },
+    {
+        "BriefDescription": "This metric estimates fraction of cycles the CPU performance was potentially limited due to Core computation issues (non divider-related).  Two distinct categories can be attributed into this metric: (1) heavy data-dependency among contiguous instructions would manifest in this metric - such cases are often referred to as low Instruction Level Parallelism (ILP). (2) Contention on some hardware execution unit other than Divider. For example; when there are too many multiply operations.",
+        "MetricExpr": "100 * ( ( ( ( CYCLE_ACTIVITY.STALLS_TOTAL + UOPS_EXECUTED.CYCLES_GE_1_UOP_EXEC - ( UOPS_EXECUTED.CYCLES_GE_3_UOPS_EXEC if ( ( INST_RETIRED.ANY / ( CPU_CLK_UNHALTED.THREAD ) ) > 1.8 ) else UOPS_EXECUTED.CYCLES_GE_2_UOPS_EXEC ) - ( RS_EVENTS.EMPTY_CYCLES if ( ( ( 4 ) * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) > 0.1 ) else 0 ) + RESOURCE_STALLS.SB ) ) - RESOURCE_STALLS.SB - CYCLE_ACTIVITY.STALLS_MEM_ANY ) / ( CPU_CLK_UNHALTED.THREAD ) )",
+        "MetricGroup": "PortsUtil;TmaL3;m_tma_core_bound_percent",
+        "MetricName": "tma_ports_utilization_percent",
+        "ScaleUnit": "1%"
+    },
+    {
+        "BriefDescription": "This category represents fraction of slots utilized by useful work i.e. issued uops that eventually get retired. Ideally; all pipeline slots would be attributed to the Retiring category.  Retiring of 100% would indicate the maximum Pipeline_Width throughput was achieved.  Maximizing Retiring typically increases the Instructions-per-cycle (see IPC metric). Note that a high Retiring value does not necessary mean there is no room for more performance.  For example; Heavy-operations or Microcode Assists are categorized under Retiring. They often indicate suboptimal performance and can often be optimized or avoided. ",
+        "MetricExpr": "100 * ( ( UOPS_RETIRED.RETIRE_SLOTS ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) )",
+        "MetricGroup": "TmaL1",
+        "MetricName": "tma_retiring_percent",
+        "ScaleUnit": "1%"
+    },
+    {
+        "BriefDescription": "This metric represents fraction of slots where the CPU was retiring light-weight operations -- instructions that require no more than one uop (micro-operation). This correlates with total number of instructions used by the program. A uops-per-instruction (see UPI metric) ratio of 1 or less should be expected for decently optimized software running on Intel Core/Xeon products. While this often indicates efficient X86 instructions were executed; high value does not necessarily mean better performance cannot be achieved.",
+        "MetricExpr": "100 * ( ( ( UOPS_RETIRED.RETIRE_SLOTS ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) - ( ( ( ( UOPS_RETIRED.RETIRE_SLOTS ) / UOPS_ISSUED.ANY ) * IDQ.MS_UOPS / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) ) )",
+        "MetricGroup": "Retire;TmaL2;m_tma_retiring_percent",
+        "MetricName": "tma_light_operations_percent",
+        "ScaleUnit": "1%"
+    },
+    {
+        "BriefDescription": "This metric represents overall arithmetic floating-point (FP) operations fraction the CPU has executed (retired). Note this metric's value may exceed its parent due to use of \"Uops\" CountDomain and FMA double-counting.",
+        "MetricExpr": "100 * ( ( INST_RETIRED.X87 * ( ( UOPS_RETIRED.RETIRE_SLOTS ) / INST_RETIRED.ANY ) / ( UOPS_RETIRED.RETIRE_SLOTS ) ) + ( ( FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE ) / ( UOPS_RETIRED.RETIRE_SLOTS ) ) + ( min( ( ( FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE ) / ( UOPS_RETIRED.RETIRE_SLOTS ) ) , ( 1 ) ) ) )",
+        "MetricGroup": "HPC;TmaL3;m_tma_light_operations_percent",
+        "MetricName": "tma_fp_arith_percent",
+        "ScaleUnit": "1%"
+    },
+    {
+        "BriefDescription": "This metric represents fraction of slots where the CPU was retiring heavy-weight operations -- instructions that require two or more uops or microcoded sequences. This highly-correlates with the uop length of these instructions/sequences.",
+        "MetricExpr": "100 * ( ( ( ( UOPS_RETIRED.RETIRE_SLOTS ) / UOPS_ISSUED.ANY ) * IDQ.MS_UOPS / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) )",
+        "MetricGroup": "Retire;TmaL2;m_tma_retiring_percent",
+        "MetricName": "tma_heavy_operations_percent",
+        "ScaleUnit": "1%"
+    },
+    {
+        "BriefDescription": "This metric represents fraction of slots the CPU was retiring uops fetched by the Microcode Sequencer (MS) unit.  The MS is used for CISC instructions not supported by the default decoders (like repeat move strings; or CPUID); or by microcode assists used to address some operation modes (like in Floating Point assists). These cases can often be avoided.",
+        "MetricExpr": "100 * ( ( ( UOPS_RETIRED.RETIRE_SLOTS ) / UOPS_ISSUED.ANY ) * IDQ.MS_UOPS / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) )",
+        "MetricGroup": "MicroSeq;TmaL3;m_tma_heavy_operations_percent",
+        "MetricName": "tma_microcode_sequencer_percent",
+        "ScaleUnit": "1%"
      }
  ]
tools/perf/pmu-events/arch/x86/broadwellx/cache.json +5 -17
index 127abe08362f..2efc4c0ee740 100644
@@ -814,9 +814,8 @@
          "EventCode": "0xB7, 0xBB",
          "EventName": "OFFCORE_RESPONSE.ALL_CODE_RD.LLC_HIT.HIT_OTHER_CORE_NO_FWD",
          "MSRIndex": "0x1a6,0x1a7",
-        "MSRValue": "0x04003C0244",
+        "MSRValue": "0x4003C0244",
          "Offcore": "1",
-        "PublicDescription": "Counts all demand & prefetch code reads hit in the L3 and the snoops to sibling cores hit in either E/S state and the line is not forwarded",
          "SampleAfterValue": "100003",
          "UMask": "0x1"
      },
@@ -829,7 +828,6 @@
          "MSRIndex": "0x1a6,0x1a7",
          "MSRValue": "0x10003C0091",
          "Offcore": "1",
-        "PublicDescription": "Counts all demand & prefetch data reads hit in the L3 and the snoop to one of the sibling cores hits the line in M state and the line is forwarded",
          "SampleAfterValue": "100003",
          "UMask": "0x1"
      },
@@ -840,9 +838,8 @@
          "EventCode": "0xB7, 0xBB",
          "EventName": "OFFCORE_RESPONSE.ALL_DATA_RD.LLC_HIT.HIT_OTHER_CORE_NO_FWD",
          "MSRIndex": "0x1a6,0x1a7",
-        "MSRValue": "0x04003C0091",
+        "MSRValue": "0x4003C0091",
          "Offcore": "1",
-        "PublicDescription": "Counts all demand & prefetch data reads hit in the L3 and the snoops to sibling cores hit in either E/S state and the line is not forwarded",
          "SampleAfterValue": "100003",
          "UMask": "0x1"
      },
@@ -855,7 +852,6 @@
          "MSRIndex": "0x1a6,0x1a7",
          "MSRValue": "0x10003C07F7",
          "Offcore": "1",
-        "PublicDescription": "Counts all data/code/rfo reads (demand & prefetch) hit in the L3 and the snoop to one of the sibling cores hits the line in M state and the line is forwarded",
          "SampleAfterValue": "100003",
          "UMask": "0x1"
      },
@@ -866,9 +862,8 @@
          "EventCode": "0xB7, 0xBB",
          "EventName": "OFFCORE_RESPONSE.ALL_READS.LLC_HIT.HIT_OTHER_CORE_NO_FWD",
          "MSRIndex": "0x1a6,0x1a7",
-        "MSRValue": "0x04003C07F7",
+        "MSRValue": "0x4003C07F7",
          "Offcore": "1",
-        "PublicDescription": "Counts all data/code/rfo reads (demand & prefetch) hit in the L3 and the snoops to sibling cores hit in either E/S state and the line is not forwarded",
          "SampleAfterValue": "100003",
          "UMask": "0x1"
      },
@@ -881,7 +876,6 @@
          "MSRIndex": "0x1a6,0x1a7",
          "MSRValue": "0x3F803C8FFF",
          "Offcore": "1",
-        "PublicDescription": "Counts all requests hit in the L3",
          "SampleAfterValue": "100003",
          "UMask": "0x1"
      },
@@ -894,7 +888,6 @@
          "MSRIndex": "0x1a6,0x1a7",
          "MSRValue": "0x10003C0122",
          "Offcore": "1",
-        "PublicDescription": "Counts all demand & prefetch RFOs hit in the L3 and the snoop to one of the sibling cores hits the line in M state and the line is forwarded",
          "SampleAfterValue": "100003",
          "UMask": "0x1"
      },
@@ -905,9 +898,8 @@
          "EventCode": "0xB7, 0xBB",
          "EventName": "OFFCORE_RESPONSE.ALL_RFO.LLC_HIT.HIT_OTHER_CORE_NO_FWD",
          "MSRIndex": "0x1a6,0x1a7",
-        "MSRValue": "0x04003C0122",
+        "MSRValue": "0x4003C0122",
          "Offcore": "1",
-        "PublicDescription": "Counts all demand & prefetch RFOs hit in the L3 and the snoops to sibling cores hit in either E/S state and the line is not forwarded",
          "SampleAfterValue": "100003",
          "UMask": "0x1"
      },
@@ -920,7 +912,6 @@
          "MSRIndex": "0x1a6,0x1a7",
          "MSRValue": "0x3F803C0002",
          "Offcore": "1",
-        "PublicDescription": "Counts all demand data writes (RFOs) hit in the L3",
          "SampleAfterValue": "100003",
          "UMask": "0x1"
      },
@@ -933,7 +924,6 @@
          "MSRIndex": "0x1a6,0x1a7",
          "MSRValue": "0x10003C0002",
          "Offcore": "1",
-        "PublicDescription": "Counts all demand data writes (RFOs) hit in the L3 and the snoop to one of the sibling cores hits the line in M state and the line is forwarded",
          "SampleAfterValue": "100003",
          "UMask": "0x1"
      },
@@ -946,7 +936,6 @@
          "MSRIndex": "0x1a6,0x1a7",
          "MSRValue": "0x3F803C0200",
          "Offcore": "1",
-        "PublicDescription": "Counts prefetch (that bring data to LLC only) code reads hit in the L3",
          "SampleAfterValue": "100003",
          "UMask": "0x1"
      },
@@ -959,7 +948,6 @@
          "MSRIndex": "0x1a6,0x1a7",
          "MSRValue": "0x3F803C0100",
          "Offcore": "1",
-        "PublicDescription": "Counts all prefetch (that bring data to LLC only) RFOs hit in the L3",
          "SampleAfterValue": "100003",
          "UMask": "0x1"
      },
@@ -973,4 +961,4 @@
          "SampleAfterValue": "100003",
          "UMask": "0x10"
      }
-]
\ No newline at end of file
+]
tools/perf/pmu-events/arch/x86/broadwellx/floating-point.json +8 -1
index 9ad37dddb354..93bbc8600321 100644
@@ -5,6 +5,7 @@
          "CounterHTOff": "0,1,2,3",
          "EventCode": "0xc7",
          "EventName": "FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE",
+        "PublicDescription": "Number of SSE/AVX computational 128-bit packed double precision floating-point instructions retired; some instructions will count twice as noted below.  Each count represents 2 computation operations, one for each element.  Applies to SSE* and AVX* packed double precision floating-point instructions: ADD SUB HADD HSUB SUBADD MUL DIV MIN MAX SQRT DPP FM(N)ADD/SUB.  DPP and FM(N)ADD/SUB instructions count twice as they perform 2 calculations per element. The DAZ and FTZ flags in the MXCSR register need to be set when using these events.",
          "SampleAfterValue": "2000003",
          "UMask": "0x4"
      },
@@ -14,6 +15,7 @@
          "CounterHTOff": "0,1,2,3",
          "EventCode": "0xc7",
          "EventName": "FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE",
+        "PublicDescription": "Number of SSE/AVX computational 128-bit packed single precision floating-point instructions retired; some instructions will count twice as noted below.  Each count represents 4 computation operations, one for each element.  Applies to SSE* and AVX* packed single precision floating-point instructions: ADD SUB HADD HSUB SUBADD MUL DIV MIN MAX SQRT RSQRT RCP DPP FM(N)ADD/SUB.  DPP and FM(N)ADD/SUB instructions count twice as they perform 2 calculations per element. The DAZ and FTZ flags in the MXCSR register need to be set when using these events.",
          "SampleAfterValue": "2000003",
          "UMask": "0x8"
      },
@@ -23,6 +25,7 @@
          "CounterHTOff": "0,1,2,3",
          "EventCode": "0xc7",
          "EventName": "FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE",
+        "PublicDescription": "Number of SSE/AVX computational 256-bit packed double precision floating-point instructions retired; some instructions will count twice as noted below.  Each count represents 4 computation operations, one for each element.  Applies to SSE* and AVX* packed double precision floating-point instructions: ADD SUB HADD HSUB SUBADD MUL DIV MIN MAX SQRT FM(N)ADD/SUB.  FM(N)ADD/SUB instructions count twice as they perform 2 calculations per element. The DAZ and FTZ flags in the MXCSR register need to be set when using these events.",
          "SampleAfterValue": "2000003",
          "UMask": "0x10"
      },
@@ -32,6 +35,7 @@
          "CounterHTOff": "0,1,2,3",
          "EventCode": "0xc7",
          "EventName": "FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE",
+        "PublicDescription": "Number of SSE/AVX computational 256-bit packed single precision floating-point instructions retired; some instructions will count twice as noted below.  Each count represents 8 computation operations, one for each element.  Applies to SSE* and AVX* packed single precision floating-point instructions: ADD SUB HADD HSUB SUBADD MUL DIV MIN MAX SQRT RSQRT RCP DPP FM(N)ADD/SUB.  DPP and FM(N)ADD/SUB instructions count twice as they perform 2 calculations per element. The DAZ and FTZ flags in the MXCSR register need to be set when using these events.",
          "SampleAfterValue": "2000003",
          "UMask": "0x20"
      },
@@ -59,6 +63,7 @@
          "CounterHTOff": "0,1,2,3",
          "EventCode": "0xc7",
          "EventName": "FP_ARITH_INST_RETIRED.SCALAR",
+        "PublicDescription": "Number of SSE/AVX computational scalar single precision and double precision floating-point instructions retired; some instructions will count twice as noted below.  Each count represents 1 computational operation. Applies to SSE* and AVX* scalar single precision floating-point instructions: ADD SUB MUL DIV MIN MAX SQRT RSQRT RCP FM(N)ADD/SUB.  FM(N)ADD/SUB instructions count twice as they perform 2 calculations per element. The DAZ and FTZ flags in the MXCSR register need to be set when using these events.",
          "SampleAfterValue": "2000003",
          "UMask": "0x3"
      },
@@ -68,6 +73,7 @@
          "CounterHTOff": "0,1,2,3",
          "EventCode": "0xc7",
          "EventName": "FP_ARITH_INST_RETIRED.SCALAR_DOUBLE",
+        "PublicDescription": "Number of SSE/AVX computational scalar double precision floating-point instructions retired; some instructions will count twice as noted below.  Each count represents 1 computational operation. Applies to SSE* and AVX* scalar double precision floating-point instructions: ADD SUB MUL DIV MIN MAX SQRT FM(N)ADD/SUB.  FM(N)ADD/SUB instructions count twice as they perform 2 calculations per element. The DAZ and FTZ flags in the MXCSR register need to be set when using these events.",
          "SampleAfterValue": "2000003",
          "UMask": "0x1"
      },
@@ -77,6 +83,7 @@
          "CounterHTOff": "0,1,2,3",
          "EventCode": "0xc7",
          "EventName": "FP_ARITH_INST_RETIRED.SCALAR_SINGLE",
+        "PublicDescription": "Number of SSE/AVX computational scalar single precision floating-point instructions retired; some instructions will count twice as noted below.  Each count represents 1 computational operation. Applies to SSE* and AVX* scalar single precision floating-point instructions: ADD SUB MUL DIV MIN MAX SQRT RSQRT RCP FM(N)ADD/SUB.  FM(N)ADD/SUB instructions count twice as they perform 2 calculations per element. The DAZ and FTZ flags in the MXCSR register need to be set when using these events.",
          "SampleAfterValue": "2000003",
          "UMask": "0x2"
      },
@@ -190,4 +197,4 @@
          "SampleAfterValue": "2000003",
          "UMask": "0x3"
      }
-]
\ No newline at end of file
+]
tools/perf/pmu-events/arch/x86/broadwellx/frontend.json +1 -1
index f0bcb945ff76..37ce8034b2ed 100644
@@ -292,4 +292,4 @@
          "SampleAfterValue": "2000003",
          "UMask": "0x1"
      }
-]
\ No newline at end of file
+]
tools/perf/pmu-events/arch/x86/broadwellx/memory.json +10 -29
index cce993b197e3..545f61f691b9 100644
@@ -247,7 +247,6 @@
          "MSRIndex": "0x1a6,0x1a7",
          "MSRValue": "0x3FBFC00244",
          "Offcore": "1",
-        "PublicDescription": "Counts all demand & prefetch code reads miss in the L3",
          "SampleAfterValue": "100003",
          "UMask": "0x1"
      },
@@ -258,9 +257,8 @@
          "EventCode": "0xB7, 0xBB",
          "EventName": "OFFCORE_RESPONSE.ALL_CODE_RD.LLC_MISS.LOCAL_DRAM",
          "MSRIndex": "0x1a6,0x1a7",
-        "MSRValue": "0x0604000244",
+        "MSRValue": "0x604000244",
          "Offcore": "1",
-        "PublicDescription": "Counts all demand & prefetch code reads miss the L3 and the data is returned from local dram",
          "SampleAfterValue": "100003",
          "UMask": "0x1"
      },
@@ -273,7 +271,6 @@
          "MSRIndex": "0x1a6,0x1a7",
          "MSRValue": "0x3FBFC00091",
          "Offcore": "1",
-        "PublicDescription": "Counts all demand & prefetch data reads miss in the L3",
          "SampleAfterValue": "100003",
          "UMask": "0x1"
      },
@@ -284,9 +281,8 @@
          "EventCode": "0xB7, 0xBB",
          "EventName": "OFFCORE_RESPONSE.ALL_DATA_RD.LLC_MISS.LOCAL_DRAM",
          "MSRIndex": "0x1a6,0x1a7",
-        "MSRValue": "0x0604000091",
+        "MSRValue": "0x604000091",
          "Offcore": "1",
-        "PublicDescription": "Counts all demand & prefetch data reads miss the L3 and the data is returned from local dram",
          "SampleAfterValue": "100003",
          "UMask": "0x1"
      },
@@ -297,9 +293,8 @@
          "EventCode": "0xB7, 0xBB",
          "EventName": "OFFCORE_RESPONSE.ALL_DATA_RD.LLC_MISS.REMOTE_DRAM",
          "MSRIndex": "0x1a6,0x1a7",
-        "MSRValue": "0x063BC00091",
+        "MSRValue": "0x63BC00091",
          "Offcore": "1",
-        "PublicDescription": "Counts all demand & prefetch data reads miss the L3 and the data is returned from remote dram",
          "SampleAfterValue": "100003",
          "UMask": "0x1"
      },
@@ -312,7 +307,6 @@
          "MSRIndex": "0x1a6,0x1a7",
          "MSRValue": "0x103FC00091",
          "Offcore": "1",
-        "PublicDescription": "Counts all demand & prefetch data reads miss the L3 and the modified data is transferred from remote cache",
          "SampleAfterValue": "100003",
          "UMask": "0x1"
      },
@@ -323,9 +317,8 @@
          "EventCode": "0xB7, 0xBB",
          "EventName": "OFFCORE_RESPONSE.ALL_DATA_RD.LLC_MISS.REMOTE_HIT_FORWARD",
          "MSRIndex": "0x1a6,0x1a7",
-        "MSRValue": "0x087FC00091",
+        "MSRValue": "0x87FC00091",
          "Offcore": "1",
-        "PublicDescription": "Counts all demand & prefetch data reads miss the L3 and clean or shared data is transferred from remote cache",
          "SampleAfterValue": "100003",
          "UMask": "0x1"
      },
@@ -338,20 +331,18 @@
          "MSRIndex": "0x1a6,0x1a7",
          "MSRValue": "0x3FBFC007F7",
          "Offcore": "1",
-        "PublicDescription": "Counts all data/code/rfo reads (demand & prefetch) miss in the L3",
          "SampleAfterValue": "100003",
          "UMask": "0x1"
      },
      {
-        "BriefDescription": "Counts all data/code/rfo reads (demand & prefetch)miss the L3 and the data is returned from local dram",
+        "BriefDescription": "Counts all data/code/rfo reads (demand & prefetch) miss the L3 and the data is returned from local dram",
          "Counter": "0,1,2,3",
          "CounterHTOff": "0,1,2,3",
          "EventCode": "0xB7, 0xBB",
          "EventName": "OFFCORE_RESPONSE.ALL_READS.LLC_MISS.LOCAL_DRAM",
          "MSRIndex": "0x1a6,0x1a7",
-        "MSRValue": "0x06040007F7",
+        "MSRValue": "0x6040007F7",
          "Offcore": "1",
-        "PublicDescription": "Counts all data/code/rfo reads (demand & prefetch)miss the L3 and the data is returned from local dram",
          "SampleAfterValue": "100003",
          "UMask": "0x1"
      },
@@ -362,9 +353,8 @@
          "EventCode": "0xB7, 0xBB",
          "EventName": "OFFCORE_RESPONSE.ALL_READS.LLC_MISS.REMOTE_DRAM",
          "MSRIndex": "0x1a6,0x1a7",
-        "MSRValue": "0x063BC007F7",
+        "MSRValue": "0x63BC007F7",
          "Offcore": "1",
-        "PublicDescription": "Counts all data/code/rfo reads (demand & prefetch) miss the L3 and the data is returned from remote dram",
          "SampleAfterValue": "100003",
          "UMask": "0x1"
      },
@@ -377,7 +367,6 @@
          "MSRIndex": "0x1a6,0x1a7",
          "MSRValue": "0x103FC007F7",
          "Offcore": "1",
-        "PublicDescription": "Counts all data/code/rfo reads (demand & prefetch) miss the L3 and the modified data is transferred from remote cache",
          "SampleAfterValue": "100003",
          "UMask": "0x1"
      },
@@ -388,9 +377,8 @@
          "EventCode": "0xB7, 0xBB",
          "EventName": "OFFCORE_RESPONSE.ALL_READS.LLC_MISS.REMOTE_HIT_FORWARD",
          "MSRIndex": "0x1a6,0x1a7",
-        "MSRValue": "0x087FC007F7",
+        "MSRValue": "0x87FC007F7",
          "Offcore": "1",
-        "PublicDescription": "Counts all data/code/rfo reads (demand & prefetch) miss the L3 and clean or shared data is transferred from remote cache",
          "SampleAfterValue": "100003",
          "UMask": "0x1"
      },
@@ -403,7 +391,6 @@
          "MSRIndex": "0x1a6,0x1a7",
          "MSRValue": "0x3FBFC08FFF",
          "Offcore": "1",
-        "PublicDescription": "Counts all requests miss in the L3",
          "SampleAfterValue": "100003",
          "UMask": "0x1"
      },
@@ -416,7 +403,6 @@
          "MSRIndex": "0x1a6,0x1a7",
          "MSRValue": "0x3FBFC00122",
          "Offcore": "1",
-        "PublicDescription": "Counts all demand & prefetch RFOs miss in the L3",
          "SampleAfterValue": "100003",
          "UMask": "0x1"
      },
@@ -427,9 +413,8 @@
          "EventCode": "0xB7, 0xBB",
          "EventName": "OFFCORE_RESPONSE.ALL_RFO.LLC_MISS.LOCAL_DRAM",
          "MSRIndex": "0x1a6,0x1a7",
-        "MSRValue": "0x0604000122",
+        "MSRValue": "0x604000122",
          "Offcore": "1",
-        "PublicDescription": "Counts all demand & prefetch RFOs miss the L3 and the data is returned from local dram",
          "SampleAfterValue": "100003",
          "UMask": "0x1"
      },
@@ -442,7 +427,6 @@
          "MSRIndex": "0x1a6,0x1a7",
          "MSRValue": "0x3FBFC00002",
          "Offcore": "1",
-        "PublicDescription": "Counts all demand data writes (RFOs) miss in the L3",
          "SampleAfterValue": "100003",
          "UMask": "0x1"
      },
@@ -455,7 +439,6 @@
          "MSRIndex": "0x1a6,0x1a7",
          "MSRValue": "0x103FC00002",
          "Offcore": "1",
-        "PublicDescription": "Counts all demand data writes (RFOs) miss the L3 and the modified data is transferred from remote cache",
          "SampleAfterValue": "100003",
          "UMask": "0x1"
      },
@@ -468,7 +451,6 @@
          "MSRIndex": "0x1a6,0x1a7",
          "MSRValue": "0x3FBFC00200",
          "Offcore": "1",
-        "PublicDescription": "Counts prefetch (that bring data to LLC only) code reads miss in the L3",
          "SampleAfterValue": "100003",
          "UMask": "0x1"
      },
@@ -481,7 +463,6 @@
          "MSRIndex": "0x1a6,0x1a7",
          "MSRValue": "0x3FBFC00100",
          "Offcore": "1",
-        "PublicDescription": "Counts all prefetch (that bring data to LLC only) RFOs miss in the L3",
          "SampleAfterValue": "100003",
          "UMask": "0x1"
      },
@@ -684,4 +665,4 @@
          "SampleAfterValue": "2000003",
          "UMask": "0x40"
      }
-]
\ No newline at end of file
+]
tools/perf/pmu-events/arch/x86/broadwellx/other.json +1 -1
index 4b360fe96698..917d145d5227 100644
@@ -41,4 +41,4 @@
          "SampleAfterValue": "2000003",
          "UMask": "0x1"
      }
-]
\ No newline at end of file
+]
tools/perf/pmu-events/arch/x86/broadwellx/pipeline.json +2 -2
index 18d21b94a4b9..f0f30081d683 100644
@@ -1369,7 +1369,7 @@
          "BriefDescription": "Cycles with less than 10 actually retired uops.",
          "Counter": "0,1,2,3",
          "CounterHTOff": "0,1,2,3",
-        "CounterMask": "10",
+        "CounterMask": "16",
          "EventCode": "0xC2",
          "EventName": "UOPS_RETIRED.TOTAL_CYCLES",
          "Invert": "1",
@@ -1377,4 +1377,4 @@
          "SampleAfterValue": "2000003",
          "UMask": "0x1"
      }
-]
\ No newline at end of file
+]
tools/perf/pmu-events/arch/x86/broadwellx/uncore-cache.json +3637 -151
index 58ed6d33d1f4..1b9c1570aa47 100644
@@ -1,11 +1,62 @@
  [
      {
-        "BriefDescription": "Uncore cache clock ticks",
+        "BriefDescription": "Bounce Control",
+        "Counter": "0,1,2,3",
+        "EventCode": "0xA",
+        "EventName": "UNC_C_BOUNCE_CONTROL",
+        "PerPkg": "1",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "Uncore Clocks",
          "Counter": "0,1,2,3",
          "EventName": "UNC_C_CLOCKTICKS",
          "PerPkg": "1",
          "Unit": "CBO"
      },
+    {
+        "BriefDescription": "Counter 0 Occupancy",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x1F",
+        "EventName": "UNC_C_COUNTER0_OCCUPANCY",
+        "PerPkg": "1",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "FaST wire asserted",
+        "Counter": "0,1",
+        "EventCode": "0x9",
+        "EventName": "UNC_C_FAST_ASSERTED",
+        "PerPkg": "1",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "Cache Lookups; Data Read Request",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x34",
+        "EventName": "UNC_C_LLC_LOOKUP.DATA_READ",
+        "PerPkg": "1",
+        "UMask": "0x3",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "Cache Lookups; Write Requests",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x34",
+        "EventName": "UNC_C_LLC_LOOKUP.WRITE",
+        "PerPkg": "1",
+        "UMask": "0x5",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "Cache Lookups; External Snoop Request",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x34",
+        "EventName": "UNC_C_LLC_LOOKUP.REMOTE_SNOOP",
+        "PerPkg": "1",
+        "UMask": "0x9",
+        "Unit": "CBO"
+    },
      {
          "BriefDescription": "All LLC Misses (code+ data rd + data wr - including demand and prefetch)",
          "Counter": "0,1,2,3",
@@ -17,6 +68,24 @@
          "UMask": "0x11",
          "Unit": "CBO"
      },
+    {
+        "BriefDescription": "Cache Lookups; Lookups that Match NID",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x34",
+        "EventName": "UNC_C_LLC_LOOKUP.NID",
+        "PerPkg": "1",
+        "UMask": "0x41",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "Cache Lookups; Any Read Request",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x34",
+        "EventName": "UNC_C_LLC_LOOKUP.READ",
+        "PerPkg": "1",
+        "UMask": "0x21",
+        "Unit": "CBO"
+    },
      {
          "BriefDescription": "M line evictions from LLC (writebacks to memory)",
          "Counter": "0,1,2,3",
@@ -28,290 +97,3707 @@
          "Unit": "CBO"
      },
      {
-        "BriefDescription": "LLC misses - demand and prefetch data reads - excludes LLC prefetches. Derived from unc_c_tor_inserts.miss_opcode",
+        "BriefDescription": "Lines Victimized; Lines in E state",
          "Counter": "0,1,2,3",
-        "EventCode": "0x35",
-        "EventName": "LLC_MISSES.DATA_READ",
-        "Filter": "filter_opc=0x182",
+        "EventCode": "0x37",
+        "EventName": "UNC_C_LLC_VICTIMS.E_STATE",
          "PerPkg": "1",
-        "ScaleUnit": "64Bytes",
-        "UMask": "0x3",
+        "UMask": "0x2",
          "Unit": "CBO"
      },
      {
-        "BriefDescription": "LLC misses - Uncacheable reads (from cpu) . Derived from unc_c_tor_inserts.miss_opcode",
+        "BriefDescription": "Lines Victimized; Lines in S State",
          "Counter": "0,1,2,3",
-        "EventCode": "0x35",
-        "EventName": "LLC_MISSES.UNCACHEABLE",
-        "Filter": "filter_opc=0x187",
+        "EventCode": "0x37",
+        "EventName": "UNC_C_LLC_VICTIMS.I_STATE",
          "PerPkg": "1",
-        "ScaleUnit": "64Bytes",
-        "UMask": "0x3",
+        "UMask": "0x4",
          "Unit": "CBO"
      },
      {
-        "BriefDescription": "MMIO reads. Derived from unc_c_tor_inserts.miss_opcode",
+        "BriefDescription": "Lines Victimized",
          "Counter": "0,1,2,3",
-        "EventCode": "0x35",
-        "EventName": "LLC_MISSES.MMIO_READ",
-        "Filter": "filter_opc=0x187,filter_nc=1",
+        "EventCode": "0x37",
+        "EventName": "UNC_C_LLC_VICTIMS.F_STATE",
          "PerPkg": "1",
-        "ScaleUnit": "64Bytes",
-        "UMask": "0x3",
+        "UMask": "0x8",
          "Unit": "CBO"
      },
      {
-        "BriefDescription": "MMIO writes. Derived from unc_c_tor_inserts.miss_opcode",
+        "BriefDescription": "Lines Victimized; Victimized Lines that Match NID",
          "Counter": "0,1,2,3",
-        "EventCode": "0x35",
-        "EventName": "LLC_MISSES.MMIO_WRITE",
-        "Filter": "filter_opc=0x18f,filter_nc=1",
+        "EventCode": "0x37",
+        "EventName": "UNC_C_LLC_VICTIMS.NID",
          "PerPkg": "1",
-        "ScaleUnit": "64Bytes",
-        "UMask": "0x3",
+        "UMask": "0x40",
          "Unit": "CBO"
      },
      {
-        "BriefDescription": "LLC prefetch misses for RFO. Derived from unc_c_tor_inserts.miss_opcode",
+        "BriefDescription": "Lines Victimized",
          "Counter": "0,1,2,3",
-        "EventCode": "0x35",
-        "EventName": "LLC_MISSES.RFO_LLC_PREFETCH",
-        "Filter": "filter_opc=0x190",
+        "EventCode": "0x37",
+        "EventName": "UNC_C_LLC_VICTIMS.MISS",
          "PerPkg": "1",
-        "ScaleUnit": "64Bytes",
-        "UMask": "0x3",
+        "UMask": "0x10",
          "Unit": "CBO"
      },
      {
-        "BriefDescription": "LLC prefetch misses for code reads. Derived from unc_c_tor_inserts.miss_opcode",
+        "BriefDescription": "Cbo Misc; Silent Snoop Eviction",
          "Counter": "0,1,2,3",
-        "EventCode": "0x35",
-        "EventName": "LLC_MISSES.CODE_LLC_PREFETCH",
-        "Filter": "filter_opc=0x191",
+        "EventCode": "0x39",
+        "EventName": "UNC_C_MISC.RSPI_WAS_FSE",
          "PerPkg": "1",
-        "ScaleUnit": "64Bytes",
-        "UMask": "0x3",
+        "UMask": "0x1",
          "Unit": "CBO"
      },
      {
-        "BriefDescription": "LLC prefetch misses for data reads. Derived from unc_c_tor_inserts.miss_opcode",
+        "BriefDescription": "Cbo Misc; Write Combining Aliasing",
          "Counter": "0,1,2,3",
-        "EventCode": "0x35",
-        "EventName": "LLC_MISSES.DATA_LLC_PREFETCH",
-        "Filter": "filter_opc=0x192",
+        "EventCode": "0x39",
+        "EventName": "UNC_C_MISC.WC_ALIASING",
          "PerPkg": "1",
-        "ScaleUnit": "64Bytes",
-        "UMask": "0x3",
+        "UMask": "0x2",
          "Unit": "CBO"
      },
      {
-        "BriefDescription": "LLC misses for PCIe read current. Derived from unc_c_tor_inserts.miss_opcode",
+        "BriefDescription": "Cbo Misc",
          "Counter": "0,1,2,3",
-        "EventCode": "0x35",
-        "EventName": "LLC_MISSES.PCIE_READ",
-        "Filter": "filter_opc=0x19e",
+        "EventCode": "0x39",
+        "EventName": "UNC_C_MISC.STARTED",
          "PerPkg": "1",
-        "ScaleUnit": "64Bytes",
-        "UMask": "0x3",
+        "UMask": "0x4",
          "Unit": "CBO"
      },
      {
-        "BriefDescription": "ItoM write misses (as part of fast string memcpy stores) + PCIe full line writes. Derived from unc_c_tor_inserts.miss_opcode",
+        "BriefDescription": "Cbo Misc; RFO HitS",
          "Counter": "0,1,2,3",
-        "EventCode": "0x35",
-        "EventName": "LLC_MISSES.PCIE_WRITE",
-        "Filter": "filter_opc=0x1c8",
+        "EventCode": "0x39",
+        "EventName": "UNC_C_MISC.RFO_HIT_S",
          "PerPkg": "1",
-        "ScaleUnit": "64Bytes",
-        "UMask": "0x3",
+        "UMask": "0x8",
          "Unit": "CBO"
      },
      {
-        "BriefDescription": "PCIe write misses (full cache line). Derived from unc_c_tor_inserts.miss_opcode",
+        "BriefDescription": "Cbo Misc; Clean Victim with raw CV=0",
          "Counter": "0,1,2,3",
-        "EventCode": "0x35",
-        "EventName": "LLC_MISSES.PCIE_NON_SNOOP_WRITE",
-        "Filter": "filter_opc=0x1c8,filter_tid=0x3e",
+        "EventCode": "0x39",
+        "EventName": "UNC_C_MISC.CVZERO_PREFETCH_VICTIM",
          "PerPkg": "1",
-        "ScaleUnit": "64Bytes",
-        "UMask": "0x3",
+        "UMask": "0x10",
          "Unit": "CBO"
      },
      {
-        "BriefDescription": "PCIe writes (partial cache line). Derived from unc_c_tor_inserts.opcode",
+        "BriefDescription": "Cbo Misc; DRd hitting non-M with raw CV=0",
          "Counter": "0,1,2,3",
-        "EventCode": "0x35",
-        "EventName": "LLC_REFERENCES.PCIE_NS_PARTIAL_WRITE",
-        "Filter": "filter_opc=0x180,filter_tid=0x3e",
+        "EventCode": "0x39",
+        "EventName": "UNC_C_MISC.CVZERO_PREFETCH_MISS",
          "PerPkg": "1",
-        "UMask": "0x1",
+        "UMask": "0x20",
          "Unit": "CBO"
      },
      {
-        "BriefDescription": "L2 demand and L2 prefetch code references to LLC. Derived from unc_c_tor_inserts.opcode",
+        "BriefDescription": "AD Ring In Use; Up and Even",
          "Counter": "0,1,2,3",
-        "EventCode": "0x35",
-        "EventName": "LLC_REFERENCES.CODE_LLC_PREFETCH",
-        "Filter": "filter_opc=0x181",
+        "EventCode": "0x1B",
+        "EventName": "UNC_C_RING_AD_USED.UP_EVEN",
          "PerPkg": "1",
-        "ScaleUnit": "64Bytes",
          "UMask": "0x1",
          "Unit": "CBO"
      },
      {
-        "BriefDescription": "Streaming stores (full cache line). Derived from unc_c_tor_inserts.opcode",
+        "BriefDescription": "AD Ring In Use; Up and Odd",
          "Counter": "0,1,2,3",
-        "EventCode": "0x35",
-        "EventName": "LLC_REFERENCES.STREAMING_FULL",
-        "Filter": "filter_opc=0x18c",
+        "EventCode": "0x1B",
+        "EventName": "UNC_C_RING_AD_USED.UP_ODD",
          "PerPkg": "1",
-        "ScaleUnit": "64Bytes",
-        "UMask": "0x1",
+        "UMask": "0x2",
          "Unit": "CBO"
      },
      {
-        "BriefDescription": "Streaming stores (partial cache line). Derived from unc_c_tor_inserts.opcode",
+        "BriefDescription": "AD Ring In Use; Down and Even",
          "Counter": "0,1,2,3",
-        "EventCode": "0x35",
-        "EventName": "LLC_REFERENCES.STREAMING_PARTIAL",
-        "Filter": "filter_opc=0x18d",
+        "EventCode": "0x1B",
+        "EventName": "UNC_C_RING_AD_USED.DOWN_EVEN",
          "PerPkg": "1",
-        "ScaleUnit": "64Bytes",
-        "UMask": "0x1",
+        "UMask": "0x4",
          "Unit": "CBO"
      },
      {
-        "BriefDescription": "PCIe read current. Derived from unc_c_tor_inserts.opcode",
+        "BriefDescription": "AD Ring In Use; Down and Odd",
          "Counter": "0,1,2,3",
-        "EventCode": "0x35",
-        "EventName": "LLC_REFERENCES.PCIE_READ",
-        "Filter": "filter_opc=0x19e",
+        "EventCode": "0x1B",
+        "EventName": "UNC_C_RING_AD_USED.DOWN_ODD",
          "PerPkg": "1",
-        "ScaleUnit": "64Bytes",
-        "UMask": "0x1",
+        "UMask": "0x8",
          "Unit": "CBO"
      },
      {
-        "BriefDescription": "PCIe write references (full cache line). Derived from unc_c_tor_inserts.opcode",
+        "BriefDescription": "AD Ring In Use; Up",
          "Counter": "0,1,2,3",
-        "EventCode": "0x35",
-        "EventName": "LLC_REFERENCES.PCIE_WRITE",
-        "Filter": "filter_opc=0x1c8,filter_tid=0x3e",
+        "EventCode": "0x1B",
+        "EventName": "UNC_C_RING_AD_USED.UP",
          "PerPkg": "1",
-        "ScaleUnit": "64Bytes",
-        "UMask": "0x1",
+        "UMask": "0x3",
          "Unit": "CBO"
      },
      {
-        "BriefDescription": "Occupancy counter for LLC data reads (demand and L2 prefetch). Derived from unc_c_tor_occupancy.miss_opcode",
-        "EventCode": "0x36",
-        "EventName": "UNC_C_TOR_OCCUPANCY.LLC_DATA_READ",
-        "Filter": "filter_opc=0x182",
+        "BriefDescription": "AD Ring In Use; Down",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x1B",
+        "EventName": "UNC_C_RING_AD_USED.DOWN",
          "PerPkg": "1",
-        "UMask": "0x3",
+        "UMask": "0xC",
          "Unit": "CBO"
      },
      {
-        "BriefDescription": "read requests to home agent",
+        "BriefDescription": "AD Ring In Use; All",
          "Counter": "0,1,2,3",
-        "EventCode": "0x1",
-        "EventName": "UNC_H_REQUESTS.READS",
+        "EventCode": "0x1B",
+        "EventName": "UNC_C_RING_AD_USED.ALL",
          "PerPkg": "1",
-        "UMask": "0x3",
-        "Unit": "HA"
+        "UMask": "0xF",
+        "Unit": "CBO"
      },
      {
-        "BriefDescription": "read requests to local home agent",
+        "BriefDescription": "AK Ring In Use; Up and Even",
          "Counter": "0,1,2,3",
-        "EventCode": "0x1",
-        "EventName": "UNC_H_REQUESTS.READS_LOCAL",
+        "EventCode": "0x1C",
+        "EventName": "UNC_C_RING_AK_USED.UP_EVEN",
          "PerPkg": "1",
          "UMask": "0x1",
-        "Unit": "HA"
+        "Unit": "CBO"
      },
      {
-        "BriefDescription": "read requests to remote home agent",
+        "BriefDescription": "AK Ring In Use; Up and Odd",
          "Counter": "0,1,2,3",
-        "EventCode": "0x1",
-        "EventName": "UNC_H_REQUESTS.READS_REMOTE",
+        "EventCode": "0x1C",
+        "EventName": "UNC_C_RING_AK_USED.UP_ODD",
          "PerPkg": "1",
          "UMask": "0x2",
-        "Unit": "HA"
+        "Unit": "CBO"
      },
      {
-        "BriefDescription": "write requests to home agent",
+        "BriefDescription": "AK Ring In Use; Down and Even",
          "Counter": "0,1,2,3",
-        "EventCode": "0x1",
-        "EventName": "UNC_H_REQUESTS.WRITES",
+        "EventCode": "0x1C",
+        "EventName": "UNC_C_RING_AK_USED.DOWN_EVEN",
+        "PerPkg": "1",
+        "UMask": "0x4",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "AK Ring In Use; Down and Odd",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x1C",
+        "EventName": "UNC_C_RING_AK_USED.DOWN_ODD",
+        "PerPkg": "1",
+        "UMask": "0x8",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "AK Ring In Use; Up",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x1C",
+        "EventName": "UNC_C_RING_AK_USED.UP",
+        "PerPkg": "1",
+        "UMask": "0x3",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "AK Ring In Use; Down",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x1C",
+        "EventName": "UNC_C_RING_AK_USED.DOWN",
          "PerPkg": "1",
          "UMask": "0xC",
-        "Unit": "HA"
+        "Unit": "CBO"
      },
      {
-        "BriefDescription": "write requests to local home agent",
+        "BriefDescription": "AK Ring In Use; All",
          "Counter": "0,1,2,3",
-        "EventCode": "0x1",
-        "EventName": "UNC_H_REQUESTS.WRITES_LOCAL",
+        "EventCode": "0x1C",
+        "EventName": "UNC_C_RING_AK_USED.ALL",
+        "PerPkg": "1",
+        "UMask": "0xF",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "BL Ring in Use; Up and Even",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x1D",
+        "EventName": "UNC_C_RING_BL_USED.UP_EVEN",
+        "PerPkg": "1",
+        "UMask": "0x1",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "BL Ring in Use; Up and Odd",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x1D",
+        "EventName": "UNC_C_RING_BL_USED.UP_ODD",
+        "PerPkg": "1",
+        "UMask": "0x2",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "BL Ring in Use; Down and Even",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x1D",
+        "EventName": "UNC_C_RING_BL_USED.DOWN_EVEN",
          "PerPkg": "1",
          "UMask": "0x4",
-        "Unit": "HA"
+        "Unit": "CBO"
      },
      {
-        "BriefDescription": "write requests to remote home agent",
+        "BriefDescription": "BL Ring in Use; Down and Odd",
          "Counter": "0,1,2,3",
-        "EventCode": "0x1",
-        "EventName": "UNC_H_REQUESTS.WRITES_REMOTE",
+        "EventCode": "0x1D",
+        "EventName": "UNC_C_RING_BL_USED.DOWN_ODD",
          "PerPkg": "1",
          "UMask": "0x8",
-        "Unit": "HA"
+        "Unit": "CBO"
      },
      {
-        "BriefDescription": "Conflict requests (requests for same address from multiple agents simultaneously)",
+        "BriefDescription": "BL Ring in Use; Up",
          "Counter": "0,1,2,3",
-        "EventCode": "0x21",
-        "EventName": "UNC_H_SNOOP_RESP.RSPCNFLCT",
+        "EventCode": "0x1D",
+        "EventName": "UNC_C_RING_BL_USED.UP",
          "PerPkg": "1",
-        "UMask": "0x40",
-        "Unit": "HA"
+        "UMask": "0x3",
+        "Unit": "CBO"
      },
      {
-        "BriefDescription": "M line forwarded from remote cache along with writeback to memory",
+        "BriefDescription": "BL Ring in Use; Down",
          "Counter": "0,1,2,3",
-        "EventCode": "0x21",
-        "EventName": "UNC_H_SNOOP_RESP.RSP_FWD_WB",
+        "EventCode": "0x1D",
+        "EventName": "UNC_C_RING_BL_USED.DOWN",
          "PerPkg": "1",
-        "ScaleUnit": "64Bytes",
-        "UMask": "0x20",
-        "Unit": "HA"
+        "UMask": "0xC",
+        "Unit": "CBO"
      },
      {
-        "BriefDescription": "M line forwarded from remote cache with no writeback to memory",
+        "BriefDescription": "BL Ring in Use; Down",
          "Counter": "0,1,2,3",
-        "EventCode": "0x21",
-        "EventName": "UNC_H_SNOOP_RESP.RSPIFWD",
+        "EventCode": "0x1D",
+        "EventName": "UNC_C_RING_BL_USED.ALL",
          "PerPkg": "1",
-        "ScaleUnit": "64Bytes",
-        "UMask": "0x4",
-        "Unit": "HA"
+        "UMask": "0xF",
+        "Unit": "CBO"
      },
      {
-        "BriefDescription": "Shared line response from remote cache",
+        "BriefDescription": "Number of LLC responses that bounced on the Ring.; AD",
          "Counter": "0,1,2,3",
-        "EventCode": "0x21",
-        "EventName": "UNC_H_SNOOP_RESP.RSPS",
+        "EventCode": "0x5",
+        "EventName": "UNC_C_RING_BOUNCES.AD",
          "PerPkg": "1",
-        "ScaleUnit": "64Bytes",
-        "UMask": "0x2",
-        "Unit": "HA"
+        "UMask": "0x1",
+        "Unit": "CBO"
      },
      {
-        "BriefDescription": "Shared line forwarded from remote cache",
+        "BriefDescription": "Number of LLC responses that bounced on the Ring.; AK",
          "Counter": "0,1,2,3",
-        "EventCode": "0x21",
-        "EventName": "UNC_H_SNOOP_RESP.RSPSFWD",
+        "EventCode": "0x5",
+        "EventName": "UNC_C_RING_BOUNCES.AK",
+        "PerPkg": "1",
+        "UMask": "0x2",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "Number of LLC responses that bounced on the Ring.; BL",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x5",
+        "EventName": "UNC_C_RING_BOUNCES.BL",
+        "PerPkg": "1",
+        "UMask": "0x4",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "Number of LLC responses that bounced on the Ring.; Snoops of processor's cache",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x5",
+        "EventName": "UNC_C_RING_BOUNCES.IV",
+        "PerPkg": "1",
+        "UMask": "0x10",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "BL Ring in Use; Any",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x1E",
+        "EventName": "UNC_C_RING_IV_USED.ANY",
+        "PerPkg": "1",
+        "UMask": "0xF",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "BL Ring in Use; Any",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x1E",
+        "EventName": "UNC_C_RING_IV_USED.UP",
+        "PerPkg": "1",
+        "UMask": "0x3",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "BL Ring in Use; Down",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x1E",
+        "EventName": "UNC_C_RING_IV_USED.DOWN",
+        "PerPkg": "1",
+        "UMask": "0xCC",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "BL Ring in Use; Any",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x1E",
+        "EventName": "UNC_C_RING_IV_USED.DN",
+        "PerPkg": "1",
+        "UMask": "0xC",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "Number of cycles the Cbo is actively throttling traffic onto the Ring in order to limit bounce traffic",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x7",
+        "EventName": "UNC_C_RING_SRC_THRTL",
+        "PerPkg": "1",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "Ingress Arbiter Blocking Cycles; IPQ",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x12",
+        "EventName": "UNC_C_RxR_EXT_STARVED.IRQ",
+        "PerPkg": "1",
+        "UMask": "0x1",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "Ingress Arbiter Blocking Cycles; IRQ",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x12",
+        "EventName": "UNC_C_RxR_EXT_STARVED.IPQ",
+        "PerPkg": "1",
+        "UMask": "0x2",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "Ingress Arbiter Blocking Cycles; PRQ",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x12",
+        "EventName": "UNC_C_RxR_EXT_STARVED.PRQ",
+        "PerPkg": "1",
+        "UMask": "0x4",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "Ingress Arbiter Blocking Cycles; ISMQ_BID",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x12",
+        "EventName": "UNC_C_RxR_EXT_STARVED.ISMQ_BIDS",
+        "PerPkg": "1",
+        "UMask": "0x8",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "Ingress Allocations; IRQ",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x13",
+        "EventName": "UNC_C_RxR_INSERTS.IRQ",
+        "PerPkg": "1",
+        "UMask": "0x1",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "Ingress Allocations; IRQ Rejected",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x13",
+        "EventName": "UNC_C_RxR_INSERTS.IRQ_REJ",
+        "PerPkg": "1",
+        "UMask": "0x2",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "Ingress Allocations; IPQ",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x13",
+        "EventName": "UNC_C_RxR_INSERTS.IPQ",
+        "PerPkg": "1",
+        "UMask": "0x4",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "Ingress Allocations; PRQ",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x13",
+        "EventName": "UNC_C_RxR_INSERTS.PRQ",
+        "PerPkg": "1",
+        "UMask": "0x10",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "Ingress Allocations; PRQ",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x13",
+        "EventName": "UNC_C_RxR_INSERTS.PRQ_REJ",
+        "PerPkg": "1",
+        "UMask": "0x20",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "Probe Queue Retries; Any Reject",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x31",
+        "EventName": "UNC_C_RxR_IPQ_RETRY.ANY",
+        "PerPkg": "1",
+        "UMask": "0x1",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "Probe Queue Retries; No Egress Credits",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x31",
+        "EventName": "UNC_C_RxR_IPQ_RETRY.FULL",
+        "PerPkg": "1",
+        "UMask": "0x2",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "Probe Queue Retries; Address Conflict",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x31",
+        "EventName": "UNC_C_RxR_IPQ_RETRY.ADDR_CONFLICT",
+        "PerPkg": "1",
+        "UMask": "0x4",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "Probe Queue Retries; No QPI Credits",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x31",
+        "EventName": "UNC_C_RxR_IPQ_RETRY.QPI_CREDITS",
+        "PerPkg": "1",
+        "UMask": "0x10",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "Probe Queue Retries; No AD Sbo Credits",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x28",
+        "EventName": "UNC_C_RxR_IPQ_RETRY2.AD_SBO",
+        "PerPkg": "1",
+        "UMask": "0x1",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "Probe Queue Retries; Target Node Filter",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x28",
+        "EventName": "UNC_C_RxR_IPQ_RETRY2.TARGET",
+        "PerPkg": "1",
+        "UMask": "0x40",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "Ingress Request Queue Rejects; Any Reject",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x32",
+        "EventName": "UNC_C_RxR_IRQ_RETRY.ANY",
+        "PerPkg": "1",
+        "UMask": "0x1",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "Ingress Request Queue Rejects; No Egress Credits",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x32",
+        "EventName": "UNC_C_RxR_IRQ_RETRY.FULL",
+        "PerPkg": "1",
+        "UMask": "0x2",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "Ingress Request Queue Rejects; Address Conflict",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x32",
+        "EventName": "UNC_C_RxR_IRQ_RETRY.ADDR_CONFLICT",
+        "PerPkg": "1",
+        "UMask": "0x4",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "Ingress Request Queue Rejects; No RTIDs",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x32",
+        "EventName": "UNC_C_RxR_IRQ_RETRY.RTID",
+        "PerPkg": "1",
+        "UMask": "0x8",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "Ingress Request Queue Rejects; No QPI Credits",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x32",
+        "EventName": "UNC_C_RxR_IRQ_RETRY.QPI_CREDITS",
+        "PerPkg": "1",
+        "UMask": "0x10",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "Ingress Request Queue Rejects; No IIO Credits",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x32",
+        "EventName": "UNC_C_RxR_IRQ_RETRY.IIO_CREDITS",
+        "PerPkg": "1",
+        "UMask": "0x20",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "Ingress Request Queue Rejects",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x32",
+        "EventName": "UNC_C_RxR_IRQ_RETRY.NID",
+        "PerPkg": "1",
+        "UMask": "0x40",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "Ingress Request Queue Rejects; No AD Sbo Credits",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x29",
+        "EventName": "UNC_C_RxR_IRQ_RETRY2.AD_SBO",
+        "PerPkg": "1",
+        "UMask": "0x1",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "Ingress Request Queue Rejects; No BL Sbo Credits",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x29",
+        "EventName": "UNC_C_RxR_IRQ_RETRY2.BL_SBO",
+        "PerPkg": "1",
+        "UMask": "0x2",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "Ingress Request Queue Rejects; Target Node Filter",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x29",
+        "EventName": "UNC_C_RxR_IRQ_RETRY2.TARGET",
+        "PerPkg": "1",
+        "UMask": "0x40",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "ISMQ Retries; Any Reject",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x33",
+        "EventName": "UNC_C_RxR_ISMQ_RETRY.ANY",
+        "PerPkg": "1",
+        "UMask": "0x1",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "ISMQ Retries; No Egress Credits",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x33",
+        "EventName": "UNC_C_RxR_ISMQ_RETRY.FULL",
+        "PerPkg": "1",
+        "UMask": "0x2",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "ISMQ Retries; No RTIDs",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x33",
+        "EventName": "UNC_C_RxR_ISMQ_RETRY.RTID",
+        "PerPkg": "1",
+        "UMask": "0x8",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "ISMQ Retries; No QPI Credits",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x33",
+        "EventName": "UNC_C_RxR_ISMQ_RETRY.QPI_CREDITS",
+        "PerPkg": "1",
+        "UMask": "0x10",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "ISMQ Retries; No IIO Credits",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x33",
+        "EventName": "UNC_C_RxR_ISMQ_RETRY.IIO_CREDITS",
+        "PerPkg": "1",
+        "UMask": "0x20",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "ISMQ Retries",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x33",
+        "EventName": "UNC_C_RxR_ISMQ_RETRY.WB_CREDITS",
+        "PerPkg": "1",
+        "UMask": "0x80",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "ISMQ Retries",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x33",
+        "EventName": "UNC_C_RxR_ISMQ_RETRY.NID",
+        "PerPkg": "1",
+        "UMask": "0x40",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "ISMQ Request Queue Rejects; No AD Sbo Credits",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x2A",
+        "EventName": "UNC_C_RxR_ISMQ_RETRY2.AD_SBO",
+        "PerPkg": "1",
+        "UMask": "0x1",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "ISMQ Request Queue Rejects; No BL Sbo Credits",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x2A",
+        "EventName": "UNC_C_RxR_ISMQ_RETRY2.BL_SBO",
+        "PerPkg": "1",
+        "UMask": "0x2",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "ISMQ Request Queue Rejects; Target Node Filter",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x2A",
+        "EventName": "UNC_C_RxR_ISMQ_RETRY2.TARGET",
+        "PerPkg": "1",
+        "UMask": "0x40",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "Ingress Occupancy; IRQ",
+        "EventCode": "0x11",
+        "EventName": "UNC_C_RxR_OCCUPANCY.IRQ",
+        "PerPkg": "1",
+        "UMask": "0x1",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "Ingress Occupancy; IRQ Rejected",
+        "EventCode": "0x11",
+        "EventName": "UNC_C_RxR_OCCUPANCY.IRQ_REJ",
+        "PerPkg": "1",
+        "UMask": "0x2",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "Ingress Occupancy; IPQ",
+        "EventCode": "0x11",
+        "EventName": "UNC_C_RxR_OCCUPANCY.IPQ",
+        "PerPkg": "1",
+        "UMask": "0x4",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "Ingress Occupancy; PRQ Rejects",
+        "EventCode": "0x11",
+        "EventName": "UNC_C_RxR_OCCUPANCY.PRQ_REJ",
+        "PerPkg": "1",
+        "UMask": "0x20",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "SBo Credits Acquired; For AD Ring",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x3D",
+        "EventName": "UNC_C_SBO_CREDITS_ACQUIRED.AD",
+        "PerPkg": "1",
+        "UMask": "0x1",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "SBo Credits Acquired; For BL Ring",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x3D",
+        "EventName": "UNC_C_SBO_CREDITS_ACQUIRED.BL",
+        "PerPkg": "1",
+        "UMask": "0x2",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "SBo Credits Occupancy; For AD Ring",
+        "EventCode": "0x3E",
+        "EventName": "UNC_C_SBO_CREDIT_OCCUPANCY.AD",
+        "PerPkg": "1",
+        "UMask": "0x1",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "SBo Credits Occupancy; For BL Ring",
+        "EventCode": "0x3E",
+        "EventName": "UNC_C_SBO_CREDIT_OCCUPANCY.BL",
+        "PerPkg": "1",
+        "UMask": "0x2",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "PCIe writes (partial cache line). Derived from unc_c_tor_inserts.opcode",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x35",
+        "EventName": "LLC_REFERENCES.PCIE_NS_PARTIAL_WRITE",
+        "Filter": "filter_opc=0x180,filter_tid=0x3e",
+        "PerPkg": "1",
+        "UMask": "0x1",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "PCIe writes (partial cache line)",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x35",
+        "EventName": "UNC_C_TOR_INSERTS.OPCODE",
+        "Filter": "filter_opc=0x180,filter_tid=0x3e",
+        "PerPkg": "1",
+        "UMask": "0x1",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "L2 demand and L2 prefetch code references to LLC. Derived from unc_c_tor_inserts.opcode",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x35",
+        "EventName": "LLC_REFERENCES.CODE_LLC_PREFETCH",
+        "Filter": "filter_opc=0x181",
+        "PerPkg": "1",
+        "ScaleUnit": "64Bytes",
+        "UMask": "0x1",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "L2 demand and L2 prefetch code references to LLC",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x35",
+        "EventName": "UNC_C_TOR_INSERTS.OPCODE",
+        "Filter": "filter_opc=0x181",
+        "PerPkg": "1",
+        "ScaleUnit": "64Bytes",
+        "UMask": "0x1",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "Streaming stores (full cache line). Derived from unc_c_tor_inserts.opcode",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x35",
+        "EventName": "LLC_REFERENCES.STREAMING_FULL",
+        "Filter": "filter_opc=0x18c",
+        "PerPkg": "1",
+        "ScaleUnit": "64Bytes",
+        "UMask": "0x1",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "Streaming stores (full cache line)",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x35",
+        "EventName": "UNC_C_TOR_INSERTS.OPCODE",
+        "Filter": "filter_opc=0x18c",
+        "PerPkg": "1",
+        "ScaleUnit": "64Bytes",
+        "UMask": "0x1",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "Streaming stores (partial cache line). Derived from unc_c_tor_inserts.opcode",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x35",
+        "EventName": "LLC_REFERENCES.STREAMING_PARTIAL",
+        "Filter": "filter_opc=0x18d",
+        "PerPkg": "1",
+        "ScaleUnit": "64Bytes",
+        "UMask": "0x1",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "Streaming stores (partial cache line)",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x35",
+        "EventName": "UNC_C_TOR_INSERTS.OPCODE",
+        "Filter": "filter_opc=0x18d",
+        "PerPkg": "1",
+        "ScaleUnit": "64Bytes",
+        "UMask": "0x1",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "PCIe read current. Derived from unc_c_tor_inserts.opcode",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x35",
+        "EventName": "LLC_REFERENCES.PCIE_READ",
+        "Filter": "filter_opc=0x19e",
+        "PerPkg": "1",
+        "ScaleUnit": "64Bytes",
+        "UMask": "0x1",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "PCIe read current",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x35",
+        "EventName": "UNC_C_TOR_INSERTS.OPCODE",
+        "Filter": "filter_opc=0x19e",
+        "PerPkg": "1",
+        "ScaleUnit": "64Bytes",
+        "UMask": "0x1",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "PCIe write references (full cache line). Derived from unc_c_tor_inserts.opcode",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x35",
+        "EventName": "LLC_REFERENCES.PCIE_WRITE",
+        "Filter": "filter_opc=0x1c8,filter_tid=0x3e",
+        "PerPkg": "1",
+        "ScaleUnit": "64Bytes",
+        "UMask": "0x1",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "PCIe write references (full cache line)",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x35",
+        "EventName": "UNC_C_TOR_INSERTS.OPCODE",
+        "Filter": "filter_opc=0x1c8,filter_tid=0x3e",
+        "PerPkg": "1",
+        "ScaleUnit": "64Bytes",
+        "UMask": "0x1",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "TOR Inserts; Evictions",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x35",
+        "EventName": "UNC_C_TOR_INSERTS.EVICTION",
+        "PerPkg": "1",
+        "UMask": "0x4",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "TOR Inserts; All",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x35",
+        "EventName": "UNC_C_TOR_INSERTS.ALL",
+        "PerPkg": "1",
+        "UMask": "0x8",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "TOR Inserts; Writebacks",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x35",
+        "EventName": "UNC_C_TOR_INSERTS.WB",
+        "PerPkg": "1",
+        "UMask": "0x10",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "LLC misses - demand and prefetch data reads - excludes LLC prefetches. Derived from unc_c_tor_inserts.miss_opcode",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x35",
+        "EventName": "LLC_MISSES.DATA_READ",
+        "Filter": "filter_opc=0x182",
+        "PerPkg": "1",
+        "ScaleUnit": "64Bytes",
+        "UMask": "0x3",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "LLC misses - demand and prefetch data reads - excludes LLC prefetches",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x35",
+        "EventName": "UNC_C_TOR_INSERTS.MISS_OPCODE",
+        "Filter": "filter_opc=0x182",
+        "PerPkg": "1",
+        "ScaleUnit": "64Bytes",
+        "UMask": "0x3",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "LLC misses - Uncacheable reads (from cpu) . Derived from unc_c_tor_inserts.miss_opcode",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x35",
+        "EventName": "LLC_MISSES.UNCACHEABLE",
+        "Filter": "filter_opc=0x187",
+        "PerPkg": "1",
+        "ScaleUnit": "64Bytes",
+        "UMask": "0x3",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "LLC misses - Uncacheable reads (from cpu) ",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x35",
+        "EventName": "UNC_C_TOR_INSERTS.MISS_OPCODE",
+        "Filter": "filter_opc=0x187",
+        "PerPkg": "1",
+        "ScaleUnit": "64Bytes",
+        "UMask": "0x3",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "MMIO reads. Derived from unc_c_tor_inserts.miss_opcode",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x35",
+        "EventName": "LLC_MISSES.MMIO_READ",
+        "Filter": "filter_opc=0x187,filter_nc=1",
+        "PerPkg": "1",
+        "ScaleUnit": "64Bytes",
+        "UMask": "0x3",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "MMIO reads",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x35",
+        "EventName": "UNC_C_TOR_INSERTS.MISS_OPCODE",
+        "Filter": "filter_opc=0x187,filter_nc=1",
+        "PerPkg": "1",
+        "ScaleUnit": "64Bytes",
+        "UMask": "0x3",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "MMIO writes. Derived from unc_c_tor_inserts.miss_opcode",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x35",
+        "EventName": "LLC_MISSES.MMIO_WRITE",
+        "Filter": "filter_opc=0x18f,filter_nc=1",
+        "PerPkg": "1",
+        "ScaleUnit": "64Bytes",
+        "UMask": "0x3",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "MMIO writes",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x35",
+        "EventName": "UNC_C_TOR_INSERTS.MISS_OPCODE",
+        "Filter": "filter_opc=0x18f,filter_nc=1",
+        "PerPkg": "1",
+        "ScaleUnit": "64Bytes",
+        "UMask": "0x3",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "LLC prefetch misses for RFO. Derived from unc_c_tor_inserts.miss_opcode",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x35",
+        "EventName": "LLC_MISSES.RFO_LLC_PREFETCH",
+        "Filter": "filter_opc=0x190",
+        "PerPkg": "1",
+        "ScaleUnit": "64Bytes",
+        "UMask": "0x3",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "LLC prefetch misses for RFO",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x35",
+        "EventName": "UNC_C_TOR_INSERTS.MISS_OPCODE",
+        "Filter": "filter_opc=0x190",
+        "PerPkg": "1",
+        "ScaleUnit": "64Bytes",
+        "UMask": "0x3",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "LLC prefetch misses for code reads. Derived from unc_c_tor_inserts.miss_opcode",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x35",
+        "EventName": "LLC_MISSES.CODE_LLC_PREFETCH",
+        "Filter": "filter_opc=0x191",
+        "PerPkg": "1",
+        "ScaleUnit": "64Bytes",
+        "UMask": "0x3",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "LLC prefetch misses for code reads",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x35",
+        "EventName": "UNC_C_TOR_INSERTS.MISS_OPCODE",
+        "Filter": "filter_opc=0x191",
+        "PerPkg": "1",
+        "ScaleUnit": "64Bytes",
+        "UMask": "0x3",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "LLC prefetch misses for data reads. Derived from unc_c_tor_inserts.miss_opcode",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x35",
+        "EventName": "LLC_MISSES.DATA_LLC_PREFETCH",
+        "Filter": "filter_opc=0x192",
+        "PerPkg": "1",
+        "ScaleUnit": "64Bytes",
+        "UMask": "0x3",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "LLC prefetch misses for data reads",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x35",
+        "EventName": "UNC_C_TOR_INSERTS.MISS_OPCODE",
+        "Filter": "filter_opc=0x192",
+        "PerPkg": "1",
+        "ScaleUnit": "64Bytes",
+        "UMask": "0x3",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "LLC misses for PCIe read current. Derived from unc_c_tor_inserts.miss_opcode",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x35",
+        "EventName": "LLC_MISSES.PCIE_READ",
+        "Filter": "filter_opc=0x19e",
+        "PerPkg": "1",
+        "ScaleUnit": "64Bytes",
+        "UMask": "0x3",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "LLC misses for PCIe read current",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x35",
+        "EventName": "UNC_C_TOR_INSERTS.MISS_OPCODE",
+        "Filter": "filter_opc=0x19e",
+        "PerPkg": "1",
+        "ScaleUnit": "64Bytes",
+        "UMask": "0x3",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "ItoM write misses (as part of fast string memcpy stores) + PCIe full line writes. Derived from unc_c_tor_inserts.miss_opcode",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x35",
+        "EventName": "LLC_MISSES.PCIE_WRITE",
+        "Filter": "filter_opc=0x1c8",
+        "PerPkg": "1",
+        "ScaleUnit": "64Bytes",
+        "UMask": "0x3",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "ItoM write misses (as part of fast string memcpy stores) + PCIe full line writes",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x35",
+        "EventName": "UNC_C_TOR_INSERTS.MISS_OPCODE",
+        "Filter": "filter_opc=0x1c8",
+        "PerPkg": "1",
+        "ScaleUnit": "64Bytes",
+        "UMask": "0x3",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "PCIe write misses (full cache line). Derived from unc_c_tor_inserts.miss_opcode",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x35",
+        "EventName": "LLC_MISSES.PCIE_NON_SNOOP_WRITE",
+        "Filter": "filter_opc=0x1c8,filter_tid=0x3e",
+        "PerPkg": "1",
+        "ScaleUnit": "64Bytes",
+        "UMask": "0x3",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "PCIe write misses (full cache line)",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x35",
+        "EventName": "UNC_C_TOR_INSERTS.MISS_OPCODE",
+        "Filter": "filter_opc=0x1c8,filter_tid=0x3e",
+        "PerPkg": "1",
+        "ScaleUnit": "64Bytes",
+        "UMask": "0x3",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "TOR Inserts; NID and Opcode Matched",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x35",
+        "EventName": "UNC_C_TOR_INSERTS.NID_OPCODE",
+        "PerPkg": "1",
+        "UMask": "0x41",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "TOR Inserts; NID Matched Evictions",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x35",
+        "EventName": "UNC_C_TOR_INSERTS.NID_EVICTION",
+        "PerPkg": "1",
+        "UMask": "0x44",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "TOR Inserts; NID Matched",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x35",
+        "EventName": "UNC_C_TOR_INSERTS.NID_ALL",
+        "PerPkg": "1",
+        "UMask": "0x48",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "TOR Inserts; NID Matched Writebacks",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x35",
+        "EventName": "UNC_C_TOR_INSERTS.NID_WB",
+        "PerPkg": "1",
+        "UMask": "0x50",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "TOR Inserts; NID and Opcode Matched Miss",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x35",
+        "EventName": "UNC_C_TOR_INSERTS.NID_MISS_OPCODE",
+        "PerPkg": "1",
+        "UMask": "0x43",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "TOR Inserts; NID Matched Miss All",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x35",
+        "EventName": "UNC_C_TOR_INSERTS.NID_MISS_ALL",
+        "PerPkg": "1",
+        "UMask": "0x4A",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "TOR Inserts; Misses to Local Memory",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x35",
+        "EventName": "UNC_C_TOR_INSERTS.MISS_LOCAL",
+        "PerPkg": "1",
+        "UMask": "0x2A",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "TOR Inserts; Misses to Remote Memory",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x35",
+        "EventName": "UNC_C_TOR_INSERTS.MISS_REMOTE",
+        "PerPkg": "1",
+        "UMask": "0x8A",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "TOR Inserts; Local Memory",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x35",
+        "EventName": "UNC_C_TOR_INSERTS.LOCAL",
+        "PerPkg": "1",
+        "UMask": "0x28",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "TOR Inserts; Remote Memory",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x35",
+        "EventName": "UNC_C_TOR_INSERTS.REMOTE",
+        "PerPkg": "1",
+        "UMask": "0x88",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "TOR Inserts; Misses to Local Memory - Opcode Matched",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x35",
+        "EventName": "UNC_C_TOR_INSERTS.MISS_LOCAL_OPCODE",
+        "PerPkg": "1",
+        "UMask": "0x23",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "TOR Inserts; Misses to Remote Memory - Opcode Matched",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x35",
+        "EventName": "UNC_C_TOR_INSERTS.MISS_REMOTE_OPCODE",
+        "PerPkg": "1",
+        "UMask": "0x83",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "TOR Inserts; Local Memory - Opcode Matched",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x35",
+        "EventName": "UNC_C_TOR_INSERTS.LOCAL_OPCODE",
+        "PerPkg": "1",
+        "UMask": "0x21",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "TOR Inserts; Remote Memory - Opcode Matched",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x35",
+        "EventName": "UNC_C_TOR_INSERTS.REMOTE_OPCODE",
+        "PerPkg": "1",
+        "UMask": "0x81",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "TOR Occupancy; Opcode Match",
+        "EventCode": "0x36",
+        "EventName": "UNC_C_TOR_OCCUPANCY.OPCODE",
+        "PerPkg": "1",
+        "UMask": "0x1",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "TOR Occupancy; Evictions",
+        "EventCode": "0x36",
+        "EventName": "UNC_C_TOR_OCCUPANCY.EVICTION",
+        "PerPkg": "1",
+        "UMask": "0x4",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "TOR Occupancy; Any",
+        "EventCode": "0x36",
+        "EventName": "UNC_C_TOR_OCCUPANCY.ALL",
+        "PerPkg": "1",
+        "UMask": "0x8",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "Occupancy counter for LLC data reads (demand and L2 prefetch). Derived from unc_c_tor_occupancy.miss_opcode",
+        "EventCode": "0x36",
+        "EventName": "UNC_C_TOR_OCCUPANCY.LLC_DATA_READ",
+        "Filter": "filter_opc=0x182",
+        "PerPkg": "1",
+        "UMask": "0x3",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "Occupancy counter for LLC data reads (demand and L2 prefetch)",
+        "EventCode": "0x36",
+        "EventName": "UNC_C_TOR_OCCUPANCY.MISS_OPCODE",
+        "Filter": "filter_opc=0x182",
+        "PerPkg": "1",
+        "UMask": "0x3",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "TOR Occupancy; Miss All",
+        "EventCode": "0x36",
+        "EventName": "UNC_C_TOR_OCCUPANCY.MISS_ALL",
+        "PerPkg": "1",
+        "UMask": "0xA",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "TOR Occupancy; NID and Opcode Matched",
+        "EventCode": "0x36",
+        "EventName": "UNC_C_TOR_OCCUPANCY.NID_OPCODE",
+        "PerPkg": "1",
+        "UMask": "0x41",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "TOR Occupancy; NID Matched Evictions",
+        "EventCode": "0x36",
+        "EventName": "UNC_C_TOR_OCCUPANCY.NID_EVICTION",
+        "PerPkg": "1",
+        "UMask": "0x44",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "TOR Occupancy; NID Matched",
+        "EventCode": "0x36",
+        "EventName": "UNC_C_TOR_OCCUPANCY.NID_ALL",
+        "PerPkg": "1",
+        "UMask": "0x48",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "TOR Occupancy; NID and Opcode Matched Miss",
+        "EventCode": "0x36",
+        "EventName": "UNC_C_TOR_OCCUPANCY.NID_MISS_OPCODE",
+        "PerPkg": "1",
+        "UMask": "0x43",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "TOR Occupancy; NID Matched",
+        "EventCode": "0x36",
+        "EventName": "UNC_C_TOR_OCCUPANCY.NID_MISS_ALL",
+        "PerPkg": "1",
+        "UMask": "0x4A",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "TOR Occupancy",
+        "EventCode": "0x36",
+        "EventName": "UNC_C_TOR_OCCUPANCY.MISS_LOCAL",
+        "PerPkg": "1",
+        "UMask": "0x2A",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "TOR Occupancy",
+        "EventCode": "0x36",
+        "EventName": "UNC_C_TOR_OCCUPANCY.MISS_REMOTE",
+        "PerPkg": "1",
+        "UMask": "0x8A",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "TOR Occupancy",
+        "EventCode": "0x36",
+        "EventName": "UNC_C_TOR_OCCUPANCY.LOCAL",
+        "PerPkg": "1",
+        "UMask": "0x28",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "TOR Occupancy",
+        "EventCode": "0x36",
+        "EventName": "UNC_C_TOR_OCCUPANCY.REMOTE",
+        "PerPkg": "1",
+        "UMask": "0x88",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "TOR Occupancy; Misses to Local Memory - Opcode Matched",
+        "EventCode": "0x36",
+        "EventName": "UNC_C_TOR_OCCUPANCY.MISS_LOCAL_OPCODE",
+        "PerPkg": "1",
+        "UMask": "0x23",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "TOR Occupancy; Misses to Remote Memory - Opcode Matched",
+        "EventCode": "0x36",
+        "EventName": "UNC_C_TOR_OCCUPANCY.MISS_REMOTE_OPCODE",
+        "PerPkg": "1",
+        "UMask": "0x83",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "TOR Occupancy; Local Memory - Opcode Matched",
+        "EventCode": "0x36",
+        "EventName": "UNC_C_TOR_OCCUPANCY.LOCAL_OPCODE",
+        "PerPkg": "1",
+        "UMask": "0x21",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "TOR Occupancy; Remote Memory - Opcode Matched",
+        "EventCode": "0x36",
+        "EventName": "UNC_C_TOR_OCCUPANCY.REMOTE_OPCODE",
+        "PerPkg": "1",
+        "UMask": "0x81",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "TOR Occupancy; Writebacks",
+        "EventCode": "0x36",
+        "EventName": "UNC_C_TOR_OCCUPANCY.WB",
+        "PerPkg": "1",
+        "UMask": "0x10",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "TOR Occupancy; NID Matched Writebacks",
+        "EventCode": "0x36",
+        "EventName": "UNC_C_TOR_OCCUPANCY.NID_WB",
+        "PerPkg": "1",
+        "UMask": "0x50",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "Onto AD Ring",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x4",
+        "EventName": "UNC_C_TxR_ADS_USED.AD",
+        "PerPkg": "1",
+        "UMask": "0x1",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "Onto AK Ring",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x4",
+        "EventName": "UNC_C_TxR_ADS_USED.AK",
+        "PerPkg": "1",
+        "UMask": "0x2",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "Onto BL Ring",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x4",
+        "EventName": "UNC_C_TxR_ADS_USED.BL",
+        "PerPkg": "1",
+        "UMask": "0x4",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "Egress Allocations; AD - Cachebo",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x2",
+        "EventName": "UNC_C_TxR_INSERTS.AD_CACHE",
+        "PerPkg": "1",
+        "UMask": "0x1",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "Egress Allocations; AK - Cachebo",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x2",
+        "EventName": "UNC_C_TxR_INSERTS.AK_CACHE",
+        "PerPkg": "1",
+        "UMask": "0x2",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "Egress Allocations; BL - Cacheno",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x2",
+        "EventName": "UNC_C_TxR_INSERTS.BL_CACHE",
+        "PerPkg": "1",
+        "UMask": "0x4",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "Egress Allocations; IV - Cachebo",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x2",
+        "EventName": "UNC_C_TxR_INSERTS.IV_CACHE",
+        "PerPkg": "1",
+        "UMask": "0x8",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "Egress Allocations; AD - Corebo",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x2",
+        "EventName": "UNC_C_TxR_INSERTS.AD_CORE",
+        "PerPkg": "1",
+        "UMask": "0x10",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "Egress Allocations; AK - Corebo",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x2",
+        "EventName": "UNC_C_TxR_INSERTS.AK_CORE",
+        "PerPkg": "1",
+        "UMask": "0x20",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "Egress Allocations; BL - Corebo",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x2",
+        "EventName": "UNC_C_TxR_INSERTS.BL_CORE",
+        "PerPkg": "1",
+        "UMask": "0x40",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "LRU Queue; LRU Age 0",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x3C",
+        "EventName": "UNC_C_QLRU.AGE0",
+        "PerPkg": "1",
+        "UMask": "0x1",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "LRU Queue; LRU Age 1",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x3C",
+        "EventName": "UNC_C_QLRU.AGE1",
+        "PerPkg": "1",
+        "UMask": "0x2",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "LRU Queue; LRU Age 2",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x3C",
+        "EventName": "UNC_C_QLRU.AGE2",
+        "PerPkg": "1",
+        "UMask": "0x4",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "LRU Queue; LRU Age 3",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x3C",
+        "EventName": "UNC_C_QLRU.AGE3",
+        "PerPkg": "1",
+        "UMask": "0x8",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "LRU Queue; LRU Bits Decremented",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x3C",
+        "EventName": "UNC_C_QLRU.LRU_DECREMENT",
+        "PerPkg": "1",
+        "UMask": "0x10",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "LRU Queue; Non-0 Aged Victim",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x3C",
+        "EventName": "UNC_C_QLRU.VICTIM_NON_ZERO",
+        "PerPkg": "1",
+        "UMask": "0x20",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "AD",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x6",
+        "EventName": "UNC_C_RING_SINK_STARVED.AD",
+        "PerPkg": "1",
+        "UMask": "0x1",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "AK",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x6",
+        "EventName": "UNC_C_RING_SINK_STARVED.AK",
+        "PerPkg": "1",
+        "UMask": "0x2",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "IV",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x6",
+        "EventName": "UNC_C_RING_SINK_STARVED.IV",
+        "PerPkg": "1",
+        "UMask": "0x8",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "BL",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x6",
+        "EventName": "UNC_C_RING_SINK_STARVED.BL",
+        "PerPkg": "1",
+        "UMask": "0x4",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "Ingress Internal Starvation Cycles; IRQ",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x14",
+        "EventName": "UNC_C_RxR_INT_STARVED.IRQ",
+        "PerPkg": "1",
+        "UMask": "0x1",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "Ingress Internal Starvation Cycles; IPQ",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x14",
+        "EventName": "UNC_C_RxR_INT_STARVED.IPQ",
+        "PerPkg": "1",
+        "UMask": "0x4",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "Ingress Internal Starvation Cycles; ISMQ",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x14",
+        "EventName": "UNC_C_RxR_INT_STARVED.ISMQ",
+        "PerPkg": "1",
+        "UMask": "0x8",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "Ingress Internal Starvation Cycles; PRQ",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x14",
+        "EventName": "UNC_C_RxR_INT_STARVED.PRQ",
+        "PerPkg": "1",
+        "UMask": "0x10",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "Injection Starvation; Onto AK Ring",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x3",
+        "EventName": "UNC_C_TxR_STARVED.AK_BOTH",
+        "PerPkg": "1",
+        "UMask": "0x2",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "Injection Starvation; Onto BL Ring",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x3",
+        "EventName": "UNC_C_TxR_STARVED.BL_BOTH",
+        "PerPkg": "1",
+        "UMask": "0x4",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "Injection Starvation; Onto IV Ring",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x3",
+        "EventName": "UNC_C_TxR_STARVED.IV",
+        "PerPkg": "1",
+        "UMask": "0x8",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "Injection Starvation; Onto AD Ring (to core)",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x3",
+        "EventName": "UNC_C_TxR_STARVED.AD_CORE",
+        "PerPkg": "1",
+        "UMask": "0x10",
+        "Unit": "CBO"
+    },
+    {
+        "BriefDescription": "QPI Address/Opcode Match; Address & Opcode Match",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x20",
+        "EventName": "UNC_H_ADDR_OPC_MATCH.FILT",
+        "PerPkg": "1",
+        "UMask": "0x3",
+        "Unit": "HA"
+    },
+    {
+        "BriefDescription": "QPI Address/Opcode Match; Address",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x20",
+        "EventName": "UNC_H_ADDR_OPC_MATCH.ADDR",
+        "PerPkg": "1",
+        "UMask": "0x1",
+        "Unit": "HA"
+    },
+    {
+        "BriefDescription": "QPI Address/Opcode Match; Opcode",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x20",
+        "EventName": "UNC_H_ADDR_OPC_MATCH.OPC",
+        "PerPkg": "1",
+        "UMask": "0x2",
+        "Unit": "HA"
+    },
+    {
+        "BriefDescription": "QPI Address/Opcode Match; AD Opcodes",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x20",
+        "EventName": "UNC_H_ADDR_OPC_MATCH.AD",
+        "PerPkg": "1",
+        "UMask": "0x4",
+        "Unit": "HA"
+    },
+    {
+        "BriefDescription": "QPI Address/Opcode Match; BL Opcodes",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x20",
+        "EventName": "UNC_H_ADDR_OPC_MATCH.BL",
+        "PerPkg": "1",
+        "UMask": "0x8",
+        "Unit": "HA"
+    },
+    {
+        "BriefDescription": "QPI Address/Opcode Match; AK Opcodes",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x20",
+        "EventName": "UNC_H_ADDR_OPC_MATCH.AK",
+        "PerPkg": "1",
+        "UMask": "0x10",
+        "Unit": "HA"
+    },
+    {
+        "BriefDescription": "BT Cycles Not Empty",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x42",
+        "EventName": "UNC_H_BT_CYCLES_NE",
+        "PerPkg": "1",
+        "Unit": "HA"
+    },
+    {
+        "BriefDescription": "HA to iMC Bypass; Taken",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x14",
+        "EventName": "UNC_H_BYPASS_IMC.TAKEN",
+        "PerPkg": "1",
+        "UMask": "0x1",
+        "Unit": "HA"
+    },
+    {
+        "BriefDescription": "HA to iMC Bypass; Not Taken",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x14",
+        "EventName": "UNC_H_BYPASS_IMC.NOT_TAKEN",
+        "PerPkg": "1",
+        "UMask": "0x2",
+        "Unit": "HA"
+    },
+    {
+        "BriefDescription": "uclks",
+        "Counter": "0,1,2,3",
+        "EventName": "UNC_H_CLOCKTICKS",
+        "PerPkg": "1",
+        "Unit": "HA"
+    },
+    {
+        "BriefDescription": "Direct2Core Messages Sent",
+        "Counter": "0,1,2,3",
+        "EventCode": "0x11",
+        "EventName": "UNC_H_DIRECT2CORE_COUNT",
[… more lines in the raw message]

Lines of this file are not shown. View the raw message.

tools/perf/pmu-events/arch/x86/broadwellx/uncore-interconnect.json +1432 -6
index 824961318c1e..071ce45620d2 100644

Lines of this file are not shown. View the raw message.

tools/perf/pmu-events/arch/x86/broadwellx/uncore-memory.json +2835 -14
index 66eed399724c..302e956a82ed 100644

Lines of this file are not shown. View the raw message.

tools/perf/pmu-events/arch/x86/broadwellx/uncore-other.json new file +3252 -0
new file mode 100644
index 000000000000..289a726c9ac3

Lines of this file are not shown. View the raw message.

tools/perf/pmu-events/arch/x86/broadwellx/uncore-power.json +401 -36
index dd1b95655d1d..3ffb70ff573d 100644

Lines of this file are not shown. View the raw message.

tools/perf/pmu-events/arch/x86/broadwellx/virtual-memory.json +1 -1
index 818a8b132c08..6a6de8790f25 100644

Lines of this file are not shown. View the raw message.

tools/perf/pmu-events/arch/x86/mapfile.csv +1 -1
index f5a382421a60..02d3434aecf3 100644

Lines of this file are not shown. View the raw message.

The rest of this diff is not shown. View the raw message.

Signature · 1 line
Zhengjun Xing

Part of the diff is not shown here. View the raw message.

A message that is not in the archive (sent off-list or not yet received): CAP-5=fV65fiadnaAmebYS1CjxwuFy4oKxV88v6oHdVPCc=n+Ow@mail.gmail.com

#5 lore raw

Re: [PATCH v1 02/31] perf vendor events: Update Intel broadwellx

HI Arnaldo,

Quoted 13 lines · Ian Rogers

On Sun, Jul 24, 2022, 6:34 PM Xing Zhengjun <[email protected] <mailto:[email protected]>> wrote:

Hi Ian,

On 7/23/2022 6:32 AM, Ian Rogers wrote:
> Use script at:
>
https://github.com/intel/event-converter-for-linux-perf/blob/master/download_and_gen.py
<https://github.com/intel/event-converter-for-linux-perf/blob/master/download_and_gen.py>
>

Quoted 9 lines

It is better to add the event JSON file version and TMA version for the future track. For example, the event list is based on broadwellx JSON file v19, the metrics are based on TMA4.4-full.

Thanks Xing,

I'll add that in v2. I'd skipped it this time as I'd been adding it to the mapfile. I'll also break apart the tremontx to snowridgex change to match yours. I also will rebase to see if that'll fix the git am issue. Apologies in advance to everyone's inboxes :-)

Hi Arnaldo,

Except for Snowridgex, I also post SPR/ADL/HSX/BDX event list last month, Can these be merged in or do I need to update them? Thanks.

https://lore.kernel.org/all/[email protected]/
https://lore.kernel.org/all/[email protected]/

https://lore.kernel.org/all/[email protected]/
https://lore.kernel.org/all/[email protected]/
https://lore.kernel.org/all/[email protected]/
https://lore.kernel.org/all/[email protected]/

Quoted 1,771 lines

Thanks,
Ian

> to download and generate the latest events and metrics. Manually copy > the broadwellx files into perf and update mapfile.csv.
>
> Tested with 'perf test':
>   10: PMU events
    :
>   10.1: PMU event table sanity
    : Ok
>   10.2: PMU event map aliases
     : Ok
>   10.3: Parsing of PMU event table metrics
    : Ok
>   10.4: Parsing of PMU event table metrics with fake PMUs : Ok
>   90: perf all metricgroups test
    : Ok
>   91: perf all metrics test
     : Skip
>   93: perf all PMU test
     : Ok
>
> Signed-off-by: Ian Rogers <[email protected]
<mailto:[email protected]>>
> ---
>   .../arch/x86/broadwellx/bdx-metrics.json      |  570 ++- >   .../pmu-events/arch/x86/broadwellx/cache.json |   22 +- >   .../arch/x86/broadwellx/floating-point.json   |    9 +- >   .../arch/x86/broadwellx/frontend.json         |    2 +- >   .../arch/x86/broadwellx/memory.json           |   39 +- >   .../pmu-events/arch/x86/broadwellx/other.json |    2 +- >   .../arch/x86/broadwellx/pipeline.json         |    4 +- >   .../arch/x86/broadwellx/uncore-cache.json     | 3788
++++++++++++++++-
>   .../x86/broadwellx/uncore-interconnect.json   | 1438 ++++++- >   .../arch/x86/broadwellx/uncore-memory.json    | 2849 ++++++++++++- >   .../arch/x86/broadwellx/uncore-other.json     | 3252 ++++++++++++++ >   .../arch/x86/broadwellx/uncore-power.json     |  437 +- >   .../arch/x86/broadwellx/virtual-memory.json   |    2 +- >   tools/perf/pmu-events/arch/x86/mapfile.csv    |    2 +- >   14 files changed, 12103 insertions(+), 313 deletions(-) >   create mode 100644
tools/perf/pmu-events/arch/x86/broadwellx/uncore-other.json
>
> diff --git
a/tools/perf/pmu-events/arch/x86/broadwellx/bdx-metrics.json b/tools/perf/pmu-events/arch/x86/broadwellx/bdx-metrics.json
> index b055947c0afe..720ee7c9332d 100644
> --- a/tools/perf/pmu-events/arch/x86/broadwellx/bdx-metrics.json > +++ b/tools/perf/pmu-events/arch/x86/broadwellx/bdx-metrics.json > @@ -74,12 +74,6 @@
>           "MetricGroup": "Branches;Fed;FetchBW",
>           "MetricName": "UpTB"
>       },
> -    {
> -        "BriefDescription": "Cycles Per Instruction (per Logical
Processor)",
> -        "MetricExpr": "1 / (INST_RETIRED.ANY /
CPU_CLK_UNHALTED.THREAD)",
> -        "MetricGroup": "Pipeline;Mem",
> -        "MetricName": "CPI"
> -    },
>       {
>           "BriefDescription": "Per-Logical Processor actual
clocks when the Logical Processor is active.",
>           "MetricExpr": "CPU_CLK_UNHALTED.THREAD",
> @@ -130,43 +124,25 @@
>           "MetricName": "FLOPc_SMT"
>       },
>       {
> -        "BriefDescription": "Actual per-core usage of the
Floating Point execution units (regardless of the vector width)",
> +        "BriefDescription": "Actual per-core usage of the
Floating Point non-X87 execution units (regardless of precision or vector-width)",
>           "MetricExpr": "( (FP_ARITH_INST_RETIRED.SCALAR_SINGLE +
FP_ARITH_INST_RETIRED.SCALAR_DOUBLE) + (FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE) ) / ( 2 * CPU_CLK_UNHALTED.THREAD )",
>           "MetricGroup": "Cor;Flops;HPC",
>           "MetricName": "FP_Arith_Utilization",
> -        "PublicDescription": "Actual per-core usage of the
Floating Point execution units (regardless of the vector width). Values > 1 are possible due to Fused-Multiply Add (FMA) counting."
> +        "PublicDescription": "Actual per-core usage of the
Floating Point non-X87 execution units (regardless of precision or vector-width). Values > 1 are possible due to ([BDW+] Fused-Multiply Add (FMA) counting - common; [ADL+] use all of ADD/MUL/FMA in Scalar or 128/256-bit vectors - less common)."
>       },
>       {
> -        "BriefDescription": "Actual per-core usage of the
Floating Point execution units (regardless of the vector width). SMT version; use when SMT is enabled and measuring per logical CPU.",
> +        "BriefDescription": "Actual per-core usage of the
Floating Point non-X87 execution units (regardless of precision or vector-width). SMT version; use when SMT is enabled and measuring per logical CPU.",
>           "MetricExpr": "( (FP_ARITH_INST_RETIRED.SCALAR_SINGLE +
FP_ARITH_INST_RETIRED.SCALAR_DOUBLE) + (FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE) ) / ( 2 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ) )",
>           "MetricGroup": "Cor;Flops;HPC_SMT",
>           "MetricName": "FP_Arith_Utilization_SMT",
> -        "PublicDescription": "Actual per-core usage of the
Floating Point execution units (regardless of the vector width). Values > 1 are possible due to Fused-Multiply Add (FMA) counting. SMT version; use when SMT is enabled and measuring per logical CPU."
> +        "PublicDescription": "Actual per-core usage of the
Floating Point non-X87 execution units (regardless of precision or vector-width). Values > 1 are possible due to ([BDW+] Fused-Multiply Add (FMA) counting - common; [ADL+] use all of ADD/MUL/FMA in Scalar or 128/256-bit vectors - less common). SMT version; use when SMT is enabled and measuring per logical CPU."
>       },
>       {
> -        "BriefDescription": "Instruction-Level-Parallelism
(average number of uops executed when there is at least 1 uop executed)",
> +        "BriefDescription": "Instruction-Level-Parallelism
(average number of uops executed when there is execution) per-core",
>           "MetricExpr": "UOPS_EXECUTED.THREAD / ((
cpu@UOPS_EXECUTED.CORE\\,cmask\\=1@ / 2 ) if #SMT_on else UOPS_EXECUTED.CYCLES_GE_1_UOP_EXEC)",
>           "MetricGroup": "Backend;Cor;Pipeline;PortsUtil", >           "MetricName": "ILP"
>       },
> -    {
> -        "BriefDescription": "Branch Misprediction Cost: Fraction
of TMA slots wasted per non-speculative branch misprediction (retired JEClear)",
> -        "MetricExpr": " ( ((BR_MISP_RETIRED.ALL_BRANCHES / (
BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT )) * (( UOPS_ISSUED.ANY - UOPS_RETIRED.RETIRE_SLOTS + 4 * INT_MISC.RECOVERY_CYCLES ) / (4 * CPU_CLK_UNHALTED.THREAD))) + (4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) * (BR_MISP_RETIRED.ALL_BRANCHES * (12 * ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT + BACLEARS.ANY ) / CPU_CLK_UNHALTED.THREAD) / ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT + BACLEARS.ANY )) / #(4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) ) * (4 * CPU_CLK_UNHALTED.THREAD) / BR_MISP_RETIRED.ALL_BRANCHES",
> -        "MetricGroup": "Bad;BrMispredicts",
> -        "MetricName": "Branch_Misprediction_Cost"
> -    },
> -    {
> -        "BriefDescription": "Branch Misprediction Cost: Fraction
of TMA slots wasted per non-speculative branch misprediction (retired JEClear)",
> -        "MetricExpr": " ( ((BR_MISP_RETIRED.ALL_BRANCHES / (
BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT )) * (( UOPS_ISSUED.ANY - UOPS_RETIRED.RETIRE_SLOTS + 4 * ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) ) / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )))) + (4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )))
* (BR_MISP_RETIRED.ALL_BRANCHES * (12 * ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT + BACLEARS.ANY ) / CPU_CLK_UNHALTED.THREAD) / ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT + BACLEARS.ANY )) / #(4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) ) * (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )) / BR_MISP_RETIRED.ALL_BRANCHES",
> -        "MetricGroup": "Bad;BrMispredicts_SMT",
> -        "MetricName": "Branch_Misprediction_Cost_SMT"
> -    },
> -    {
> -        "BriefDescription": "Number of Instructions per
non-speculative Branch Misprediction (JEClear)",
> -        "MetricExpr": "INST_RETIRED.ANY /
BR_MISP_RETIRED.ALL_BRANCHES",
> -        "MetricGroup": "Bad;BadSpec;BrMispredicts",
> -        "MetricName": "IpMispredict"
> -    },
>       {
>           "BriefDescription": "Core actual clocks when any
Logical Processor is active on the Physical Core",
>           "MetricExpr": "( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1
+ CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )", > @@ -256,6 +232,18 @@
>           "MetricGroup": "Summary;TmaL1",
>           "MetricName": "Instructions"
>       },
> +    {
> +        "BriefDescription": "Average number of Uops retired in
cycles where at least one uop has retired.",
> +        "MetricExpr": "UOPS_RETIRED.RETIRE_SLOTS /
cpu@UOPS_RETIRED.RETIRE_SLOTS\\,cmask\\=1@",
> +        "MetricGroup": "Pipeline;Ret",
> +        "MetricName": "Retire"
> +    },
> +    {
> +        "BriefDescription": "",
> +        "MetricExpr": "UOPS_EXECUTED.THREAD /
cpu@UOPS_EXECUTED.THREAD\\,cmask\\=1@",
> +        "MetricGroup": "Cor;Pipeline;PortsUtil;SMT",
> +        "MetricName": "Execute"
> +    },
>       {
>           "BriefDescription": "Fraction of Uops delivered by the
DSB (aka Decoded ICache; or Uop Cache)",
>           "MetricExpr": "IDQ.DSB_UOPS / (( IDQ.DSB_UOPS +
LSD.UOPS + IDQ.MITE_UOPS + IDQ.MS_UOPS ) )",
> @@ -263,35 +251,34 @@
>           "MetricName": "DSB_Coverage"
>       },
>       {
> -        "BriefDescription": "Actual Average Latency for L1
data-cache miss demand load instructions (in core cycles)",
> -        "MetricExpr": "L1D_PEND_MISS.PENDING / (
MEM_LOAD_UOPS_RETIRED.L1_MISS + mem_load_uops_retired.hit_lfb )",
> -        "MetricGroup": "Mem;MemoryBound;MemoryLat",
> -        "MetricName": "Load_Miss_Real_Latency",
> -        "PublicDescription": "Actual Average Latency for L1
data-cache miss demand load instructions (in core cycles). Latency may be overestimated for multi-load instructions - e.g. repeat strings."
> +        "BriefDescription": "Number of Instructions per
non-speculative Branch Misprediction (JEClear) (lower number means higher occurrence rate)",
> +        "MetricExpr": "INST_RETIRED.ANY /
BR_MISP_RETIRED.ALL_BRANCHES",
> +        "MetricGroup": "Bad;BadSpec;BrMispredicts",
> +        "MetricName": "IpMispredict"
>       },
>       {
> -        "BriefDescription": "Memory-Level-Parallelism (average
number of L1 miss demand load when there is at least one such miss. Per-Logical Processor)",
> -        "MetricExpr": "L1D_PEND_MISS.PENDING /
L1D_PEND_MISS.PENDING_CYCLES",
> -        "MetricGroup": "Mem;MemoryBound;MemoryBW",
> -        "MetricName": "MLP"
> +        "BriefDescription": "Branch Misprediction Cost: Fraction
of TMA slots wasted per non-speculative branch misprediction (retired JEClear)",
> +        "MetricExpr": " ( ((BR_MISP_RETIRED.ALL_BRANCHES / (
BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT )) * (( UOPS_ISSUED.ANY - UOPS_RETIRED.RETIRE_SLOTS + 4 * INT_MISC.RECOVERY_CYCLES ) / (4 * CPU_CLK_UNHALTED.THREAD))) + (4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) * (BR_MISP_RETIRED.ALL_BRANCHES * (12 * ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT + BACLEARS.ANY ) / CPU_CLK_UNHALTED.THREAD) / ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT + BACLEARS.ANY )) / #(4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * CPU_CLK_UNHALTED.THREAD)) ) * (4 * CPU_CLK_UNHALTED.THREAD) / BR_MISP_RETIRED.ALL_BRANCHES",
> +        "MetricGroup": "Bad;BrMispredicts",
> +        "MetricName": "Branch_Misprediction_Cost"
>       },
>       {
> -        "BriefDescription": "Average data fill bandwidth to the
L1 data cache [GB / sec]",
> -        "MetricExpr": "64 * L1D.REPLACEMENT / 1000000000 /
duration_time",
> -        "MetricGroup": "Mem;MemoryBW",
> -        "MetricName": "L1D_Cache_Fill_BW"
> +        "BriefDescription": "Branch Misprediction Cost: Fraction
of TMA slots wasted per non-speculative branch misprediction (retired JEClear)",
> +        "MetricExpr": " ( ((BR_MISP_RETIRED.ALL_BRANCHES / (
BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT )) * (( UOPS_ISSUED.ANY - UOPS_RETIRED.RETIRE_SLOTS + 4 * ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) ) / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )))) + (4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )))
* (BR_MISP_RETIRED.ALL_BRANCHES * (12 * ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT + BACLEARS.ANY ) / CPU_CLK_UNHALTED.THREAD) / ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT + BACLEARS.ANY )) / #(4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))) ) * (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )) / BR_MISP_RETIRED.ALL_BRANCHES",
> +        "MetricGroup": "Bad;BrMispredicts_SMT",
> +        "MetricName": "Branch_Misprediction_Cost_SMT"
>       },
>       {
> -        "BriefDescription": "Average data fill bandwidth to the
L2 cache [GB / sec]",
> -        "MetricExpr": "64 * L2_LINES_IN.ALL / 1000000000 /
duration_time",
> -        "MetricGroup": "Mem;MemoryBW",
> -        "MetricName": "L2_Cache_Fill_BW"
> +        "BriefDescription": "Actual Average Latency for L1
data-cache miss demand load operations (in core cycles)",
> +        "MetricExpr": "L1D_PEND_MISS.PENDING / (
MEM_LOAD_UOPS_RETIRED.L1_MISS + mem_load_uops_retired.hit_lfb )",
> +        "MetricGroup": "Mem;MemoryBound;MemoryLat",
> +        "MetricName": "Load_Miss_Real_Latency"
>       },
>       {
> -        "BriefDescription": "Average per-core data fill
bandwidth to the L3 cache [GB / sec]",
> -        "MetricExpr": "64 * LONGEST_LAT_CACHE.MISS / 1000000000
/ duration_time",
> -        "MetricGroup": "Mem;MemoryBW",
> -        "MetricName": "L3_Cache_Fill_BW"
> +        "BriefDescription": "Memory-Level-Parallelism (average
number of L1 miss demand load when there is at least one such miss. Per-Logical Processor)",
> +        "MetricExpr": "L1D_PEND_MISS.PENDING /
L1D_PEND_MISS.PENDING_CYCLES",
> +        "MetricGroup": "Mem;MemoryBound;MemoryBW",
> +        "MetricName": "MLP"
>       },
>       {
>           "BriefDescription": "L1 cache true misses per kilo
instruction for retired demand loads",
> @@ -306,13 +293,13 @@
>           "MetricName": "L2MPKI"
>       },
>       {
> -        "BriefDescription": "L2 cache misses per kilo
instruction for all request types (including speculative)",
> +        "BriefDescription": "L2 cache ([RKL+] true) misses per
kilo instruction for all request types (including speculative)",
>           "MetricExpr": "1000 * L2_RQSTS.MISS / INST_RETIRED.ANY", >           "MetricGroup": "Mem;CacheMisses;Offcore",
>           "MetricName": "L2MPKI_All"
>       },
>       {
> -        "BriefDescription": "L2 cache misses per kilo
instruction for all demand loads  (including speculative)",
> +        "BriefDescription": "L2 cache ([RKL+] true) misses per
kilo instruction for all demand loads  (including speculative)",
>           "MetricExpr": "1000 * L2_RQSTS.DEMAND_DATA_RD_MISS /
INST_RETIRED.ANY",
>           "MetricGroup": "Mem;CacheMisses",
>           "MetricName": "L2MPKI_Load"
> @@ -348,6 +335,48 @@
>           "MetricGroup": "Mem;MemoryTLB_SMT",
>           "MetricName": "Page_Walks_Utilization_SMT"
>       },
> +    {
> +        "BriefDescription": "Average per-core data fill
bandwidth to the L1 data cache [GB / sec]",
> +        "MetricExpr": "64 * L1D.REPLACEMENT / 1000000000 /
duration_time",
> +        "MetricGroup": "Mem;MemoryBW",
> +        "MetricName": "L1D_Cache_Fill_BW"
> +    },
> +    {
> +        "BriefDescription": "Average per-core data fill
bandwidth to the L2 cache [GB / sec]",
> +        "MetricExpr": "64 * L2_LINES_IN.ALL / 1000000000 /
duration_time",
> +        "MetricGroup": "Mem;MemoryBW",
> +        "MetricName": "L2_Cache_Fill_BW"
> +    },
> +    {
> +        "BriefDescription": "Average per-core data fill
bandwidth to the L3 cache [GB / sec]",
> +        "MetricExpr": "64 * LONGEST_LAT_CACHE.MISS / 1000000000
/ duration_time",
> +        "MetricGroup": "Mem;MemoryBW",
> +        "MetricName": "L3_Cache_Fill_BW"
> +    },
> +    {
> +        "BriefDescription": "Average per-thread data fill
bandwidth to the L1 data cache [GB / sec]",
> +        "MetricExpr": "(64 * L1D.REPLACEMENT / 1000000000 /
duration_time)",
> +        "MetricGroup": "Mem;MemoryBW",
> +        "MetricName": "L1D_Cache_Fill_BW_1T"
> +    },
> +    {
> +        "BriefDescription": "Average per-thread data fill
bandwidth to the L2 cache [GB / sec]",
> +        "MetricExpr": "(64 * L2_LINES_IN.ALL / 1000000000 /
duration_time)",
> +        "MetricGroup": "Mem;MemoryBW",
> +        "MetricName": "L2_Cache_Fill_BW_1T"
> +    },
> +    {
> +        "BriefDescription": "Average per-thread data fill
bandwidth to the L3 cache [GB / sec]",
> +        "MetricExpr": "(64 * LONGEST_LAT_CACHE.MISS / 1000000000
/ duration_time)",
> +        "MetricGroup": "Mem;MemoryBW",
> +        "MetricName": "L3_Cache_Fill_BW_1T"
> +    },
> +    {
> +        "BriefDescription": "Average per-thread data access
bandwidth to the L3 cache [GB / sec]",
> +        "MetricExpr": "0",
> +        "MetricGroup": "Mem;MemoryBW;Offcore",
> +        "MetricName": "L3_Cache_Access_BW_1T"
> +    },
>       {
>           "BriefDescription": "Average CPU Utilization", >           "MetricExpr": "CPU_CLK_UNHALTED.REF_TSC / msr@tsc@", > @@ -364,7 +393,8 @@
>           "BriefDescription": "Giga Floating Point Operations Per
Second",
>           "MetricExpr": "( ( 1 * (
FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE ) + 2 * FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + 4 * ( FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE ) + 8 * FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE ) / 1000000000 ) / duration_time",
>           "MetricGroup": "Cor;Flops;HPC",
> -        "MetricName": "GFLOPs"
> +        "MetricName": "GFLOPs",
> +        "PublicDescription": "Giga Floating Point Operations Per
Second. Aggregate across all supported options of: FP precisions, scalar and vector instructions, vector-width and AMX engine."
>       },
>       {
>           "BriefDescription": "Average Frequency Utilization
relative nominal frequency",
> @@ -461,5 +491,439 @@
>           "MetricExpr": "(cstate_pkg@c7\\-residency@ / msr@tsc@)
* 100",
>           "MetricGroup": "Power",
>           "MetricName": "C7_Pkg_Residency"
> +    },
> +    {
> +        "BriefDescription": "CPU operating frequency (in GHz)", > +        "MetricExpr": "( CPU_CLK_UNHALTED.THREAD /
CPU_CLK_UNHALTED.REF_TSC * #SYSTEM_TSC_FREQ ) / 1000000000",
> +        "MetricGroup": "",
> +        "MetricName": "cpu_operating_frequency",
> +        "ScaleUnit": "1GHz"
> +    },
> +    {
> +        "BriefDescription": "Cycles per instruction retired;
indicating how much time each executed instruction took; in units of cycles.",
> +        "MetricExpr": "CPU_CLK_UNHALTED.THREAD / INST_RETIRED.ANY", > +        "MetricGroup": "",
> +        "MetricName": "cpi",
> +        "ScaleUnit": "1per_instr"
> +    },
> +    {
> +        "BriefDescription": "The ratio of number of completed
memory load instructions to the total number completed instructions",
> +        "MetricExpr": "MEM_UOPS_RETIRED.ALL_LOADS /
INST_RETIRED.ANY",
> +        "MetricGroup": "",
> +        "MetricName": "loads_per_instr",
> +        "ScaleUnit": "1per_instr"
> +    },
> +    {
> +        "BriefDescription": "The ratio of number of completed
memory store instructions to the total number completed instructions",
> +        "MetricExpr": "MEM_UOPS_RETIRED.ALL_STORES /
INST_RETIRED.ANY",
> +        "MetricGroup": "",
> +        "MetricName": "stores_per_instr",
> +        "ScaleUnit": "1per_instr"
> +    },
> +    {
> +        "BriefDescription": "Ratio of number of requests missing
L1 data cache (includes data+rfo w/ prefetches) to the total number of completed instructions",
> +        "MetricExpr": "L1D.REPLACEMENT / INST_RETIRED.ANY", > +        "MetricGroup": "",
> +        "MetricName":
"l1d_mpi_includes_data_plus_rfo_with_prefetches",
> +        "ScaleUnit": "1per_instr"
> +    },
> +    {
> +        "BriefDescription": "Ratio of number of demand load
requests hitting in L1 data cache to the total number of completed instructions",
> +        "MetricExpr": "MEM_LOAD_UOPS_RETIRED.L1_HIT /
INST_RETIRED.ANY",
> +        "MetricGroup": "",
> +        "MetricName": "l1d_demand_data_read_hits_per_instr", > +        "ScaleUnit": "1per_instr"
> +    },
> +    {
> +        "BriefDescription": "Ratio of number of code read
requests missing in L1 instruction cache (includes prefetches) to the total number of completed instructions",
> +        "MetricExpr": "L2_RQSTS.ALL_CODE_RD / INST_RETIRED.ANY", > +        "MetricGroup": "",
> +        "MetricName":
"l1_i_code_read_misses_with_prefetches_per_instr",
> +        "ScaleUnit": "1per_instr"
> +    },
> +    {
> +        "BriefDescription": "Ratio of number of completed demand
load requests hitting in L2 cache to the total number of completed instructions",
> +        "MetricExpr": "MEM_LOAD_UOPS_RETIRED.L2_HIT /
INST_RETIRED.ANY",
> +        "MetricGroup": "",
> +        "MetricName": "l2_demand_data_read_hits_per_instr", > +        "ScaleUnit": "1per_instr"
> +    },
> +    {
> +        "BriefDescription": "Ratio of number of requests missing
L2 cache (includes code+data+rfo w/ prefetches) to the total number of completed instructions",
> +        "MetricExpr": "L2_LINES_IN.ALL / INST_RETIRED.ANY", > +        "MetricGroup": "",
> +        "MetricName":
"l2_mpi_includes_code_plus_data_plus_rfo_with_prefetches",
> +        "ScaleUnit": "1per_instr"
> +    },
> +    {
> +        "BriefDescription": "Ratio of number of completed data
read request missing L2 cache to the total number of completed instructions",
> +        "MetricExpr": "MEM_LOAD_UOPS_RETIRED.L2_MISS /
INST_RETIRED.ANY",
> +        "MetricGroup": "",
> +        "MetricName": "l2_demand_data_read_mpi",
> +        "ScaleUnit": "1per_instr"
> +    },
> +    {
> +        "BriefDescription": "Ratio of number of code read
request missing L2 cache to the total number of completed instructions",
> +        "MetricExpr": "L2_RQSTS.CODE_RD_MISS / INST_RETIRED.ANY", > +        "MetricGroup": "",
> +        "MetricName": "l2_demand_code_mpi",
> +        "ScaleUnit": "1per_instr"
> +    },
> +    {
> +        "BriefDescription": "Ratio of number of data read
requests missing last level core cache (includes demand w/ prefetches) to the total number of completed instructions",
> +        "MetricExpr": "(
cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x182@ + cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x192@ ) / INST_RETIRED.ANY",
> +        "MetricGroup": "",
> +        "MetricName": "llc_data_read_mpi_demand_plus_prefetch", > +        "ScaleUnit": "1per_instr"
> +    },
> +    {
> +        "BriefDescription": "Ratio of number of code read
requests missing last level core cache (includes demand w/ prefetches) to the total number of completed instructions",
> +        "MetricExpr": "(
cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x181@ + cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x191@ ) / INST_RETIRED.ANY",
> +        "MetricGroup": "",
> +        "MetricName": "llc_code_read_mpi_demand_plus_prefetch", > +        "ScaleUnit": "1per_instr"
> +    },
> +    {
> +        "BriefDescription": "Average latency of a last level
cache (LLC) demand and prefetch data read miss (read memory access) in nano seconds",
> +        "MetricExpr": "( 1000000000 * (
cbox@UNC_C_TOR_OCCUPANCY.MISS_OPCODE\\,filter_opc\\=0x182@ / cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x182@ ) / ( UNC_C_CLOCKTICKS / ( source_count(UNC_C_CLOCKTICKS) * #num_packages ) ) ) * duration_time",
> +        "MetricGroup": "",
> +        "MetricName":
"llc_data_read_demand_plus_prefetch_miss_latency",
> +        "ScaleUnit": "1ns"
> +    },
> +    {
> +        "BriefDescription": "Average latency of a last level
cache (LLC) demand and prefetch data read miss (read memory access) addressed to local memory in nano seconds",
> +        "MetricExpr": "( 1000000000 * (
cbox@UNC_C_TOR_OCCUPANCY.MISS_LOCAL_OPCODE\\,filter_opc\\=0x182@ / cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x182@ ) / ( UNC_C_CLOCKTICKS / ( source_count(UNC_C_CLOCKTICKS) * #num_packages ) ) ) * duration_time",
> +        "MetricGroup": "",
> +        "MetricName":
"llc_data_read_demand_plus_prefetch_miss_latency_for_local_requests",
> +        "ScaleUnit": "1ns"
> +    },
> +    {
> +        "BriefDescription": "Average latency of a last level
cache (LLC) demand and prefetch data read miss (read memory access) addressed to remote memory in nano seconds",
> +        "MetricExpr": "( 1000000000 * (
cbox@UNC_C_TOR_OCCUPANCY.MISS_REMOTE_OPCODE\\,filter_opc\\=0x182@ / cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x182@ ) / ( UNC_C_CLOCKTICKS / ( source_count(UNC_C_CLOCKTICKS) * #num_packages ) ) ) * duration_time",
> +        "MetricGroup": "",
> +        "MetricName":
"llc_data_read_demand_plus_prefetch_miss_latency_for_remote_requests",
> +        "ScaleUnit": "1ns"
> +    },
> +    {
> +        "BriefDescription": "Ratio of number of completed page
walks (for all page sizes) caused by a code fetch to the total number of completed instructions. This implies it missed in the ITLB (Instruction TLB) and further levels of TLB.",
> +        "MetricExpr": "ITLB_MISSES.WALK_COMPLETED /
INST_RETIRED.ANY",
> +        "MetricGroup": "",
> +        "MetricName": "itlb_mpi",
> +        "ScaleUnit": "1per_instr"
> +    },
> +    {
> +        "BriefDescription": "Ratio of number of completed page
walks (for 2 megabyte and 4 megabyte page sizes) caused by a code fetch to the total number of completed instructions. This implies it missed in the Instruction Translation Lookaside Buffer (ITLB) and further levels of TLB.",
> +        "MetricExpr": "ITLB_MISSES.WALK_COMPLETED_2M_4M /
INST_RETIRED.ANY",
> +        "MetricGroup": "",
> +        "MetricName": "itlb_large_page_mpi",
> +        "ScaleUnit": "1per_instr"
> +    },
> +    {
> +        "BriefDescription": "Ratio of number of completed page
walks (for all page sizes) caused by demand data loads to the total number of completed instructions. This implies it missed in the DTLB and further levels of TLB.",
> +        "MetricExpr": "DTLB_LOAD_MISSES.WALK_COMPLETED /
INST_RETIRED.ANY",
> +        "MetricGroup": "",
> +        "MetricName": "dtlb_load_mpi",
> +        "ScaleUnit": "1per_instr"
> +    },
> +    {
> +        "BriefDescription": "Ratio of number of completed page
walks (for all page sizes) caused by demand data stores to the total number of completed instructions. This implies it missed in the DTLB and further levels of TLB.",
> +        "MetricExpr": "DTLB_STORE_MISSES.WALK_COMPLETED /
INST_RETIRED.ANY",
> +        "MetricGroup": "",
> +        "MetricName": "dtlb_store_mpi",
> +        "ScaleUnit": "1per_instr"
> +    },
> +    {
> +        "BriefDescription": "Memory read that miss the last
level cache (LLC) addressed to local DRAM as a percentage of total memory read accesses, does not include LLC prefetches.",
> +        "MetricExpr": "100 *
cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x182@ / ( cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x182@ + cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x182@ )",
> +        "MetricGroup": "",
> +        "MetricName": "numa_percent_reads_addressed_to_local_dram", > +        "ScaleUnit": "1%"
> +    },
> +    {
> +        "BriefDescription": "Memory reads that miss the last
level cache (LLC) addressed to remote DRAM as a percentage of total memory read accesses, does not include LLC prefetches.",
> +        "MetricExpr": "100 *
cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x182@ / ( cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x182@ + cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x182@ )",
> +        "MetricGroup": "",
> +        "MetricName": "numa_percent_reads_addressed_to_remote_dram", > +        "ScaleUnit": "1%"
> +    },
> +    {
> +        "BriefDescription": "Uncore operating frequency in GHz", > +        "MetricExpr": "UNC_C_CLOCKTICKS / (
source_count(UNC_C_CLOCKTICKS) * #num_packages ) / 1000000000",
> +        "MetricGroup": "",
> +        "MetricName": "uncore_frequency",
> +        "ScaleUnit": "1GHz"
> +    },
> +    {
> +        "BriefDescription": "Intel(R) Quick Path Interconnect
(QPI) data transmit bandwidth (MB/sec)",
> +        "MetricExpr": "( UNC_Q_TxL_FLITS_G0.DATA * 8 / 1000000)
/ duration_time",
> +        "MetricGroup": "",
> +        "MetricName": "qpi_data_transmit_bw_only_data", > +        "ScaleUnit": "1MB/s"
> +    },
> +    {
> +        "BriefDescription": "DDR memory read bandwidth (MB/sec)", > +        "MetricExpr": "( UNC_M_CAS_COUNT.RD * 64 / 1000000) /
duration_time",
> +        "MetricGroup": "",
> +        "MetricName": "memory_bandwidth_read",
> +        "ScaleUnit": "1MB/s"
> +    },
> +    {
> +        "BriefDescription": "DDR memory write bandwidth (MB/sec)", > +        "MetricExpr": "( UNC_M_CAS_COUNT.WR * 64 / 1000000) /
duration_time",
> +        "MetricGroup": "",
> +        "MetricName": "memory_bandwidth_write",
> +        "ScaleUnit": "1MB/s"
> +    },
> +    {
> +        "BriefDescription": "DDR memory bandwidth (MB/sec)", > +        "MetricExpr": "(( UNC_M_CAS_COUNT.RD +
UNC_M_CAS_COUNT.WR ) * 64 / 1000000) / duration_time",
> +        "MetricGroup": "",
> +        "MetricName": "memory_bandwidth_total",
> +        "ScaleUnit": "1MB/s"
> +    },
> +    {
> +        "BriefDescription": "Bandwidth of IO reads that are
initiated by end device controllers that are requesting memory from the CPU.",
> +        "MetricExpr": "(
cbox@UNC_C_TOR_INSERTS.OPCODE\\,filter_opc\\=0x19e@ * 64 / 1000000) / duration_time",
> +        "MetricGroup": "",
> +        "MetricName": "io_bandwidth_read",
> +        "ScaleUnit": "1MB/s"
> +    },
> +    {
> +        "BriefDescription": "Bandwidth of IO writes that are
initiated by end device controllers that are writing memory to the CPU.",
> +        "MetricExpr": "((
cbox@UNC_C_TOR_INSERTS.OPCODE\\,filter_opc\\=0x1c8\\,filter_tid\\=0x3e@ + cbox@UNC_C_TOR_INSERTS.OPCODE\\,filter_opc\\=0x180\\,filter_tid\\=0x3e@ ) * 64 / 1000000) / duration_time",
> +        "MetricGroup": "",
> +        "MetricName": "io_bandwidth_write",
> +        "ScaleUnit": "1MB/s"
> +    },
> +    {
> +        "BriefDescription": "Uops delivered from decoded
instruction cache (decoded stream buffer or DSB) as a percent of total uops delivered to Instruction Decode Queue",
> +        "MetricExpr": "100 * ( IDQ.DSB_UOPS / UOPS_ISSUED.ANY )", > +        "MetricGroup": "",
> +        "MetricName":
"percent_uops_delivered_from_decoded_icache_dsb",
> +        "ScaleUnit": "1%"
> +    },
> +    {
> +        "BriefDescription": "Uops delivered from legacy decode
pipeline (Micro-instruction Translation Engine or MITE) as a percent of total uops delivered to Instruction Decode Queue",
> +        "MetricExpr": "100 * ( IDQ.MITE_UOPS / UOPS_ISSUED.ANY )", > +        "MetricGroup": "",
> +        "MetricName":
"percent_uops_delivered_from_legacy_decode_pipeline_mite",
> +        "ScaleUnit": "1%"
> +    },
> +    {
> +        "BriefDescription": "Uops delivered from microcode
sequencer (MS) as a percent of total uops delivered to Instruction Decode Queue",
> +        "MetricExpr": "100 * ( IDQ.MS_UOPS / UOPS_ISSUED.ANY )", > +        "MetricGroup": "",
> +        "MetricName":
"percent_uops_delivered_from_microcode_sequencer_ms",
> +        "ScaleUnit": "1%"
> +    },
> +    {
> +        "BriefDescription": "Uops delivered from loop stream
detector(LSD) as a percent of total uops delivered to Instruction Decode Queue",
> +        "MetricExpr": "100 * ( LSD.UOPS / UOPS_ISSUED.ANY )", > +        "MetricGroup": "",
> +        "MetricName":
"percent_uops_delivered_from_loop_stream_detector_lsd",
> +        "ScaleUnit": "1%"
> +    },
> +    {
> +        "BriefDescription": "This category represents fraction
of slots where the processor's Frontend undersupplies its Backend. Frontend denotes the first part of the processor core responsible to fetch operations that are executed later on by the Backend part. Within the Frontend; a branch predictor predicts the next address to fetch; cache-lines are fetched from the memory subsystem; parsed into instructions; and lastly decoded into micro-operations (uops). Ideally the Frontend can issue Machine_Width uops every cycle to the Backend. Frontend Bound denotes unutilized issue-slots when there is no Backend stall; i.e. bubbles where Frontend delivered no uops while Backend could have accepted them. For example; stalls due to instruction-cache misses would be categorized under Frontend Bound.",
> +        "MetricExpr": "100 * ( IDQ_UOPS_NOT_DELIVERED.CORE / ( (
4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) )",
> +        "MetricGroup": "TmaL1;PGO",
> +        "MetricName": "tma_frontend_bound_percent",
> +        "ScaleUnit": "1%"
> +    },
> +    {
> +        "BriefDescription": "This metric represents fraction of
slots the CPU was stalled due to Frontend latency issues.  For example; instruction-cache misses; iTLB misses or fetch stalls after a branch misprediction are categorized under Frontend Latency. In such cases; the Frontend eventually delivers no uops for some period.",
> +        "MetricExpr": "100 * ( ( 4 ) *
IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) )",
> +        "MetricGroup":
"Frontend;TmaL2;m_tma_frontend_bound_percent",
> +        "MetricName": "tma_fetch_latency_percent",
> +        "ScaleUnit": "1%"
> +    },
> +    {
> +        "BriefDescription": "This metric represents fraction of
cycles the CPU was stalled due to instruction cache misses.",
> +        "MetricExpr": "100 * ( ICACHE.IFDATA_STALL / (
CPU_CLK_UNHALTED.THREAD ) )",
> +        "MetricGroup":
"BigFoot;FetchLat;IcMiss;TmaL3;m_tma_fetch_latency_percent",
> +        "MetricName": "tma_icache_misses_percent",
> +        "ScaleUnit": "1%"
> +    },
> +    {
> +        "BriefDescription": "This metric represents fraction of
cycles the CPU was stalled due to Instruction TLB (ITLB) misses.",
> +        "MetricExpr": "100 * ( ( 14 * ITLB_MISSES.STLB_HIT +
cpu@ITLB_MISSES.WALK_DURATION\\,cmask\\=0x1@ + 7 * ITLB_MISSES.WALK_COMPLETED ) / ( CPU_CLK_UNHALTED.THREAD ) )",
> +        "MetricGroup":
"BigFoot;FetchLat;MemoryTLB;TmaL3;m_tma_fetch_latency_percent",
> +        "MetricName": "tma_itlb_misses_percent",
> +        "ScaleUnit": "1%"
> +    },
> +    {
> +        "BriefDescription": "This metric represents fraction of
cycles the CPU was stalled due to Branch Resteers. Branch Resteers estimates the Frontend delay in fetching operations from corrected path; following all sorts of miss-predicted branches. For example; branchy code with lots of miss-predictions might get categorized under Branch Resteers. Note the value of this node may overlap with its siblings.",
> +        "MetricExpr": "100 * ( ( 12 ) * (
BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT + BACLEARS.ANY ) / ( CPU_CLK_UNHALTED.THREAD ) )",
> +        "MetricGroup": "FetchLat;TmaL3;m_tma_fetch_latency_percent", > +        "MetricName": "tma_branch_resteers_percent",
> +        "ScaleUnit": "1%"
> +    },
> +    {
> +        "BriefDescription": "This metric represents fraction of
cycles the CPU was stalled due to switches from DSB to MITE pipelines. The DSB (decoded i-cache) is a Uop Cache where the front-end directly delivers Uops (micro operations) avoiding heavy x86 decoding. The DSB pipeline has shorter latency and delivered higher bandwidth than the MITE (legacy instruction decode pipeline). Switching between the two pipelines can cause penalties hence this metric measures the exposed penalty.",
> +        "MetricExpr": "100 * ( DSB2MITE_SWITCHES.PENALTY_CYCLES
/ ( CPU_CLK_UNHALTED.THREAD ) )",
> +        "MetricGroup":
"DSBmiss;FetchLat;TmaL3;m_tma_fetch_latency_percent",
> +        "MetricName": "tma_dsb_switches_percent",
> +        "ScaleUnit": "1%"
> +    },
> +    {
> +        "BriefDescription": "This metric represents fraction of
cycles CPU was stalled due to Length Changing Prefixes (LCPs). Using proper compiler flags or Intel Compiler by default will certainly avoid this. #Link: Optimization Guide about LCP BKMs.",
> +        "MetricExpr": "100 * ( ILD_STALL.LCP / (
CPU_CLK_UNHALTED.THREAD ) )",
> +        "MetricGroup": "FetchLat;TmaL3;m_tma_fetch_latency_percent", > +        "MetricName": "tma_lcp_percent",
> +        "ScaleUnit": "1%"
> +    },
> +    {
> +        "BriefDescription": "This metric estimates the fraction
of cycles when the CPU was stalled due to switches of uop delivery to the Microcode Sequencer (MS). Commonly used instructions are optimized for delivery by the DSB (decoded i-cache) or MITE (legacy instruction decode) pipelines. Certain operations cannot be handled natively by the execution pipeline; and must be performed by microcode (small programs injected into the execution stream). Switching to the MS too often can negatively impact performance. The MS is designated to deliver long uop flows required by CISC instructions like CPUID; or uncommon conditions like Floating Point Assists when dealing with Denormals.",
> +        "MetricExpr": "100 * ( ( 2 ) * IDQ.MS_SWITCHES / (
CPU_CLK_UNHALTED.THREAD ) )",
> +        "MetricGroup":
"FetchLat;MicroSeq;TmaL3;m_tma_fetch_latency_percent",
> +        "MetricName": "tma_ms_switches_percent",
> +        "ScaleUnit": "1%"
> +    },
> +    {
> +        "BriefDescription": "This metric represents fraction of
slots the CPU was stalled due to Frontend bandwidth issues.  For example; inefficiencies at the instruction decoders; or restrictions for caching in the DSB (decoded uops cache) are categorized under Fetch Bandwidth. In such cases; the Frontend typically delivers suboptimal amount of uops to the Backend.",
> +        "MetricExpr": "100 * ( ( IDQ_UOPS_NOT_DELIVERED.CORE / (
( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) - ( ( 4 ) * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) )",
> +        "MetricGroup":
"FetchBW;Frontend;TmaL2;m_tma_frontend_bound_percent",
> +        "MetricName": "tma_fetch_bandwidth_percent",
> +        "ScaleUnit": "1%"
> +    },
> +    {
> +        "BriefDescription": "This metric represents Core
fraction of cycles in which CPU was likely limited due to the MITE pipeline (the legacy decode pipeline). This pipeline is used for code that was not pre-cached in the DSB or LSD. For example; inefficiencies due to asymmetric decoders; use of long immediate or LCP can manifest as MITE fetch bandwidth bottleneck.",
> +        "MetricExpr": "100 * ( ( IDQ.ALL_MITE_CYCLES_ANY_UOPS -
IDQ.ALL_MITE_CYCLES_4_UOPS ) / ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) / 2 )",
> +        "MetricGroup":
"DSBmiss;FetchBW;TmaL3;m_tma_fetch_bandwidth_percent",
> +        "MetricName": "tma_mite_percent",
> +        "ScaleUnit": "1%"
> +    },
> +    {
> +        "BriefDescription": "This metric represents Core
fraction of cycles in which CPU was likely limited due to DSB (decoded uop cache) fetch pipeline.  For example; inefficient utilization of the DSB cache structure or bank conflict when reading from it; are categorized here.",
> +        "MetricExpr": "100 * ( ( IDQ.ALL_DSB_CYCLES_ANY_UOPS -
IDQ.ALL_DSB_CYCLES_4_UOPS ) / ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) / 2 )",
> +        "MetricGroup":
"DSB;FetchBW;TmaL3;m_tma_fetch_bandwidth_percent",
> +        "MetricName": "tma_dsb_percent",
> +        "ScaleUnit": "1%"
> +    },
> +    {
> +        "BriefDescription": "This category represents fraction
of slots wasted due to incorrect speculations. This include slots used to issue uops that do not eventually get retired and slots for which the issue-pipeline was blocked due to recovery from earlier incorrect speculation. For example; wasted work due to miss-predicted branches are categorized under Bad Speculation category. Incorrect data speculation followed by Memory Ordering Nukes is another example.",
> +        "MetricExpr": "100 * ( ( UOPS_ISSUED.ANY - (
UOPS_RETIRED.RETIRE_SLOTS ) + ( 4 ) * ( ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) if #SMT_on else INT_MISC.RECOVERY_CYCLES ) ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) )",
> +        "MetricGroup": "TmaL1",
> +        "MetricName": "tma_bad_speculation_percent",
> +        "ScaleUnit": "1%"
> +    },
> +    {
> +        "BriefDescription": "This metric represents fraction of
slots the CPU has wasted due to Branch Misprediction.  These slots are either wasted by uops fetched from an incorrectly speculated program path; or stalls when the out-of-order part of the machine needs to recover its state from a speculative path.",
> +        "MetricExpr": "100 * ( ( BR_MISP_RETIRED.ALL_BRANCHES /
( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT ) ) * ( ( UOPS_ISSUED.ANY - ( UOPS_RETIRED.RETIRE_SLOTS ) + ( 4 ) * ( ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) if #SMT_on else INT_MISC.RECOVERY_CYCLES ) ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) )",
> +        "MetricGroup":
"BadSpec;BrMispredicts;TmaL2;m_tma_bad_speculation_percent",
> +        "MetricName": "tma_branch_mispredicts_percent", > +        "ScaleUnit": "1%"
> +    },
> +    {
> +        "BriefDescription": "This metric represents fraction of
slots the CPU has wasted due to Machine Clears.  These slots are either wasted by uops fetched prior to the clear; or stalls the out-of-order portion of the machine needs to recover its state after the clear. For example; this can happen due to memory ordering Nukes (e.g. Memory Disambiguation) or Self-Modifying-Code (SMC) nukes.",
> +        "MetricExpr": "100 * ( ( ( UOPS_ISSUED.ANY - (
UOPS_RETIRED.RETIRE_SLOTS ) + ( 4 ) * ( ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) if #SMT_on else INT_MISC.RECOVERY_CYCLES ) ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) - ( ( BR_MISP_RETIRED.ALL_BRANCHES / ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT ) ) * ( ( UOPS_ISSUED.ANY - ( UOPS_RETIRED.RETIRE_SLOTS ) + ( 4 ) * ( ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) if #SMT_on else INT_MISC.RECOVERY_CYCLES ) ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) ) )",
> +        "MetricGroup":
"BadSpec;MachineClears;TmaL2;m_tma_bad_speculation_percent",
> +        "MetricName": "tma_machine_clears_percent",
> +        "ScaleUnit": "1%"
> +    },
> +    {
> +        "BriefDescription": "This category represents fraction
of slots where no uops are being delivered due to a lack of required resources for accepting new uops in the Backend. Backend is the portion of the processor core where the out-of-order scheduler dispatches ready uops into their respective execution units; and once completed these uops get retired according to program order. For example; stalls due to data-cache misses or stalls due to the divider unit being overloaded are both categorized under Backend Bound. Backend Bound is further divided into two main categories:
Memory Bound and Core Bound.",
> +        "MetricExpr": "100 * ( 1 - ( (
IDQ_UOPS_NOT_DELIVERED.CORE / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) + ( ( UOPS_ISSUED.ANY - ( UOPS_RETIRED.RETIRE_SLOTS ) + ( 4 ) * ( ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) if #SMT_on else INT_MISC.RECOVERY_CYCLES ) ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) + ( ( UOPS_RETIRED.RETIRE_SLOTS ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) ) )",
> +        "MetricGroup": "TmaL1",
> +        "MetricName": "tma_backend_bound_percent",
> +        "ScaleUnit": "1%"
> +    },
> +    {
> +        "BriefDescription": "This metric represents fraction of
slots the Memory subsystem within the Backend was a bottleneck. Memory Bound estimates fraction of slots where pipeline is likely stalled due to demand load or store instructions. This accounts mainly for (1) non-completed in-flight memory demand loads which coincides with execution units starvation; in addition to (2) cases where stores could impose backpressure on the pipeline when many of them get buffered at the same time (less common out of the two).",
> +        "MetricExpr": "100 * ( ( ( CYCLE_ACTIVITY.STALLS_MEM_ANY
+ RESOURCE_STALLS.SB <http://RESOURCE_STALLS.SB> ) / ( ( CYCLE_ACTIVITY.STALLS_TOTAL + UOPS_EXECUTED.CYCLES_GE_1_UOP_EXEC - ( UOPS_EXECUTED.CYCLES_GE_3_UOPS_EXEC if ( ( INST_RETIRED.ANY / ( CPU_CLK_UNHALTED.THREAD ) ) > 1.8 ) else UOPS_EXECUTED.CYCLES_GE_2_UOPS_EXEC ) - ( RS_EVENTS.EMPTY_CYCLES if ( ( ( 4 ) * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) > 0.1 ) else 0 ) + RESOURCE_STALLS.SB <http://RESOURCE_STALLS.SB> ) ) ) * ( 1 - ( ( IDQ_UOPS_NOT_DELIVERED.CORE / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) + ( ( UOPS_ISSUED.ANY - ( UOPS_RETIRED.RETIRE_SLOTS ) + ( 4 ) * ( ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) if #SMT_on else INT_MISC.RECOVERY_CYCLES ) ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) + ( ( UOPS_RETIRED.RETIRE_SLOTS ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) ) ) )",
> +        "MetricGroup": "Backend;TmaL2;m_tma_backend_bound_percent", > +        "MetricName": "tma_memory_bound_percent",
> +        "ScaleUnit": "1%"
> +    },
> +    {
> +        "BriefDescription": "This metric estimates how often the
CPU was stalled without loads missing the L1 data cache.  The L1 data cache typically has the shortest latency.  However; in certain cases like loads blocked on older stores; a load might suffer due to high latency even though it is being satisfied by the L1. Another example is loads who miss in the TLB. These cases are characterized by execution unit stalls; while some non-completed demand load lives in the machine without having that demand load missing the L1 cache.",
> +        "MetricExpr": "100 * ( max( (
CYCLE_ACTIVITY.STALLS_MEM_ANY - CYCLE_ACTIVITY.STALLS_L1D_MISS ) / ( CPU_CLK_UNHALTED.THREAD ) , 0 ) )",
> +        "MetricGroup":
"CacheMisses;MemoryBound;TmaL3mem;TmaL3;m_tma_memory_bound_percent",
> +        "MetricName": "tma_l1_bound_percent",
> +        "ScaleUnit": "1%"
> +    },
> +    {
> +        "BriefDescription": "This metric estimates how often the
CPU was stalled due to L2 cache accesses by loads.  Avoiding cache misses (i.e. L1 misses/L2 hits) can improve the latency and increase performance.",
> +        "MetricExpr": "100 * ( ( CYCLE_ACTIVITY.STALLS_L1D_MISS
- CYCLE_ACTIVITY.STALLS_L2_MISS ) / ( CPU_CLK_UNHALTED.THREAD ) )", > +        "MetricGroup":
"CacheMisses;MemoryBound;TmaL3mem;TmaL3;m_tma_memory_bound_percent",
> +        "MetricName": "tma_l2_bound_percent",
> +        "ScaleUnit": "1%"
> +    },
> +    {
> +        "BriefDescription": "This metric estimates how often the
CPU was stalled due to loads accesses to L3 cache or contended with a sibling Core.  Avoiding cache misses (i.e. L2 misses/L3 hits) can improve the latency and increase performance.",
> +        "MetricExpr": "100 * ( ( MEM_LOAD_UOPS_RETIRED.L3_HIT /
( MEM_LOAD_UOPS_RETIRED.L3_HIT + ( 7 ) * MEM_LOAD_UOPS_RETIRED.L3_MISS ) ) * CYCLE_ACTIVITY.STALLS_L2_MISS / ( CPU_CLK_UNHALTED.THREAD ) )",
> +        "MetricGroup":
"CacheMisses;MemoryBound;TmaL3mem;TmaL3;m_tma_memory_bound_percent",
> +        "MetricName": "tma_l3_bound_percent",
> +        "ScaleUnit": "1%"
> +    },
> +    {
> +        "BriefDescription": "This metric estimates how often the
CPU was stalled on accesses to external memory (DRAM) by loads. Better caching can improve the latency and increase performance.",
> +        "MetricExpr": "100 * ( min( ( ( 1 - (
MEM_LOAD_UOPS_RETIRED.L3_HIT / ( MEM_LOAD_UOPS_RETIRED.L3_HIT + ( 7 ) * MEM_LOAD_UOPS_RETIRED.L3_MISS ) ) ) * CYCLE_ACTIVITY.STALLS_L2_MISS / ( CPU_CLK_UNHALTED.THREAD ) ) , ( 1 ) ) )",
> +        "MetricGroup":
"MemoryBound;TmaL3mem;TmaL3;m_tma_memory_bound_percent",
> +        "MetricName": "tma_dram_bound_percent",
> +        "ScaleUnit": "1%"
> +    },
> +    {
> +        "BriefDescription": "This metric estimates how often CPU
was stalled  due to RFO store memory accesses; RFO store issue a read-for-ownership request before the write. Even though store accesses do not typically stall out-of-order CPUs; there are few cases where stores can lead to actual stalls. This metric will be flagged should RFO stores be a bottleneck.",
> +        "MetricExpr": "100 * ( RESOURCE_STALLS.SB
<http://RESOURCE_STALLS.SB> / ( CPU_CLK_UNHALTED.THREAD ) )",
> +        "MetricGroup":
"MemoryBound;TmaL3mem;TmaL3;m_tma_memory_bound_percent",
> +        "MetricName": "tma_store_bound_percent",
> +        "ScaleUnit": "1%"
> +    },
> +    {
> +        "BriefDescription": "This metric represents fraction of
slots where Core non-memory issues were of a bottleneck.  Shortage in hardware compute resources; or dependencies in software's instructions are both categorized under Core Bound. Hence it may indicate the machine ran out of an out-of-order resource; certain execution units are overloaded or dependencies in program's data- or instruction-flow are limiting the performance (e.g. FP-chained long-latency arithmetic operations).",
> +        "MetricExpr": "100 * ( ( 1 - ( (
IDQ_UOPS_NOT_DELIVERED.CORE / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) + ( ( UOPS_ISSUED.ANY - ( UOPS_RETIRED.RETIRE_SLOTS ) + ( 4 ) * ( ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) if #SMT_on else INT_MISC.RECOVERY_CYCLES ) ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) + ( ( UOPS_RETIRED.RETIRE_SLOTS ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) ) ) - ( ( ( CYCLE_ACTIVITY.STALLS_MEM_ANY + RESOURCE_STALLS.SB
<http://RESOURCE_STALLS.SB> ) / ( ( CYCLE_ACTIVITY.STALLS_TOTAL + UOPS_EXECUTED.CYCLES_GE_1_UOP_EXEC - ( UOPS_EXECUTED.CYCLES_GE_3_UOPS_EXEC if ( ( INST_RETIRED.ANY / ( CPU_CLK_UNHALTED.THREAD ) ) > 1.8 ) else UOPS_EXECUTED.CYCLES_GE_2_UOPS_EXEC ) - ( RS_EVENTS.EMPTY_CYCLES if ( ( ( 4 ) * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) > 0.1 ) else 0 ) + RESOURCE_STALLS.SB <http://RESOURCE_STALLS.SB> ) ) ) * ( 1 - ( ( IDQ_UOPS_NOT_DELIVERED.CORE / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) + ( ( UOPS_ISSUED.ANY - ( UOPS_RETIRED.RETIRE_SLOTS ) + ( 4 ) * ( ( INT_MISC.RECOVERY_CYCLES_ANY / 2 ) if #SMT_on else INT_MISC.RECOVERY_CYCLES ) ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) + ( ( UOPS_RETIRED.RETIRE_SLOTS ) / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) ) ) ) )",
> +        "MetricGroup":
"Backend;TmaL2;Compute;m_tma_backend_bound_percent",
> +        "MetricName": "tma_core_bound_percent",
> +        "ScaleUnit": "1%"
> +    },
> +    {
> +        "BriefDescription": "This metric represents fraction of
cycles where the Divider unit was active. Divide and square root instructions are performed by the Divider unit and can take considerably longer latency than integer or Floating Point addition; subtraction; or multiplication.",
> +        "MetricExpr": "100 * ( ARITH.FPU_DIV_ACTIVE / ( (
CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) )",
> +        "MetricGroup": "TmaL3;m_tma_core_bound_percent", > +        "MetricName": "tma_divider_percent",
> +        "ScaleUnit": "1%"
> +    },
> +    {
> +        "BriefDescription": "This metric estimates fraction of
cycles the CPU performance was potentially limited due to Core computation issues (non divider-related).  Two distinct categories can be attributed into this metric: (1) heavy data-dependency among contiguous instructions would manifest in this metric - such cases are often referred to as low Instruction Level Parallelism (ILP).
(2) Contention on some hardware execution unit other than Divider. For example; when there are too many multiply operations.", > +        "MetricExpr": "100 * ( ( ( ( CYCLE_ACTIVITY.STALLS_TOTAL
+ UOPS_EXECUTED.CYCLES_GE_1_UOP_EXEC - ( UOPS_EXECUTED.CYCLES_GE_3_UOPS_EXEC if ( ( INST_RETIRED.ANY / ( CPU_CLK_UNHALTED.THREAD ) ) > 1.8 ) else UOPS_EXECUTED.CYCLES_GE_2_UOPS_EXEC ) - ( RS_EVENTS.EMPTY_CYCLES if ( ( ( 4 ) * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) > 0.1 ) else 0 ) + RESOURCE_STALLS.SB <http://RESOURCE_STALLS.SB> ) ) - RESOURCE_STALLS.SB <http://RESOURCE_STALLS.SB> - CYCLE_ACTIVITY.STALLS_MEM_ANY ) / ( CPU_CLK_UNHALTED.THREAD ) )", > +        "MetricGroup": "PortsUtil;TmaL3;m_tma_core_bound_percent", > +        "MetricName": "tma_ports_utilization_percent",
> +        "ScaleUnit": "1%"
> +    },
> +    {
> +        "BriefDescription": "This category represents fraction
of slots utilized by useful work i.e. issued uops that eventually get retired. Ideally; all pipeline slots would be attributed to the Retiring category.  Retiring of 100% would indicate the maximum Pipeline_Width throughput was achieved.  Maximizing Retiring typically increases the Instructions-per-cycle (see IPC metric). Note that a high Retiring value does not necessary mean there is no room for more performance.  For example; Heavy-operations or Microcode Assists are categorized under Retiring. They often indicate suboptimal performance and can often be optimized or avoided. ",
> +        "MetricExpr": "100 * ( ( UOPS_RETIRED.RETIRE_SLOTS ) / (
( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) )",
> +        "MetricGroup": "TmaL1",
> +        "MetricName": "tma_retiring_percent",
> +        "ScaleUnit": "1%"
> +    },
> +    {
> +        "BriefDescription": "This metric represents fraction of
slots where the CPU was retiring light-weight operations -- instructions that require no more than one uop (micro-operation). This correlates with total number of instructions used by the program. A uops-per-instruction (see UPI metric) ratio of 1 or less should be expected for decently optimized software running on Intel Core/Xeon products. While this often indicates efficient X86 instructions were executed; high value does not necessarily mean better performance cannot be achieved.",
> +        "MetricExpr": "100 * ( ( ( UOPS_RETIRED.RETIRE_SLOTS ) /
( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) - ( ( ( ( UOPS_RETIRED.RETIRE_SLOTS ) / UOPS_ISSUED.ANY ) * IDQ.MS_UOPS / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) ) )",
> +        "MetricGroup": "Retire;TmaL2;m_tma_retiring_percent", > +        "MetricName": "tma_light_operations_percent",
> +        "ScaleUnit": "1%"
> +    },
> +    {
> +        "BriefDescription": "This metric represents overall
arithmetic floating-point (FP) operations fraction the CPU has executed (retired). Note this metric's value may exceed its parent due to use of \"Uops\" CountDomain and FMA double-counting.",
> +        "MetricExpr": "100 * ( ( INST_RETIRED.X87 * ( (
UOPS_RETIRED.RETIRE_SLOTS ) / INST_RETIRED.ANY ) / ( UOPS_RETIRED.RETIRE_SLOTS ) ) + ( ( FP_ARITH_INST_RETIRED.SCALAR_SINGLE + FP_ARITH_INST_RETIRED.SCALAR_DOUBLE ) / ( UOPS_RETIRED.RETIRE_SLOTS ) ) + ( min( ( ( FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE + FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE + FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE ) / ( UOPS_RETIRED.RETIRE_SLOTS ) ) , ( 1 ) ) ) )",
> +        "MetricGroup": "HPC;TmaL3;m_tma_light_operations_percent", > +        "MetricName": "tma_fp_arith_percent",
> +        "ScaleUnit": "1%"
> +    },
> +    {
> +        "BriefDescription": "This metric represents fraction of
slots where the CPU was retiring heavy-weight operations -- instructions that require two or more uops or microcoded sequences. This highly-correlates with the uop length of these instructions/sequences.",
> +        "MetricExpr": "100 * ( ( ( ( UOPS_RETIRED.RETIRE_SLOTS )
/ UOPS_ISSUED.ANY ) * IDQ.MS_UOPS / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) ) )",
> +        "MetricGroup": "Retire;TmaL2;m_tma_retiring_percent", > +        "MetricName": "tma_heavy_operations_percent",
> +        "ScaleUnit": "1%"
> +    },
> +    {
> +        "BriefDescription": "This metric represents fraction of
slots the CPU was retiring uops fetched by the Microcode Sequencer (MS) unit.  The MS is used for CISC instructions not supported by the default decoders (like repeat move strings; or CPUID); or by microcode assists used to address some operation modes (like in Floating Point assists). These cases can often be avoided.",
> +        "MetricExpr": "100 * ( ( ( UOPS_RETIRED.RETIRE_SLOTS ) /
UOPS_ISSUED.ANY ) * IDQ.MS_UOPS / ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) ) )",
> +        "MetricGroup":
"MicroSeq;TmaL3;m_tma_heavy_operations_percent",
> +        "MetricName": "tma_microcode_sequencer_percent", > +        "ScaleUnit": "1%"
>       }
>   ]
> diff --git a/tools/perf/pmu-events/arch/x86/broadwellx/cache.json
b/tools/perf/pmu-events/arch/x86/broadwellx/cache.json
> index 127abe08362f..2efc4c0ee740 100644
> --- a/tools/perf/pmu-events/arch/x86/broadwellx/cache.json > +++ b/tools/perf/pmu-events/arch/x86/broadwellx/cache.json > @@ -814,9 +814,8 @@
>           "EventCode": "0xB7, 0xBB",
>           "EventName":
"OFFCORE_RESPONSE.ALL_CODE_RD.LLC_HIT.HIT_OTHER_CORE_NO_FWD",
>           "MSRIndex": "0x1a6,0x1a7",
> -        "MSRValue": "0x04003C0244",
> +        "MSRValue": "0x4003C0244",
>           "Offcore": "1",
> -        "PublicDescription": "Counts all demand & prefetch code
reads hit in the L3 and the snoops to sibling cores hit in either E/S state and the line is not forwarded",
>           "SampleAfterValue": "100003",
>           "UMask": "0x1"
>       },
> @@ -829,7 +828,6 @@
>           "MSRIndex": "0x1a6,0x1a7",
>           "MSRValue": "0x10003C0091",
>           "Offcore": "1",
> -        "PublicDescription": "Counts all demand & prefetch data
reads hit in the L3 and the snoop to one of the sibling cores hits the line in M state and the line is forwarded",
>           "SampleAfterValue": "100003",
>           "UMask": "0x1"
>       },
> @@ -840,9 +838,8 @@
>           "EventCode": "0xB7, 0xBB",
>           "EventName":
"OFFCORE_RESPONSE.ALL_DATA_RD.LLC_HIT.HIT_OTHER_CORE_NO_FWD",
>           "MSRIndex": "0x1a6,0x1a7",
> -        "MSRValue": "0x04003C0091",
> +        "MSRValue": "0x4003C0091",
>           "Offcore": "1",
> -        "PublicDescription": "Counts all demand & prefetch data
reads hit in the L3 and the snoops to sibling cores hit in either E/S state and the line is not forwarded",
>           "SampleAfterValue": "100003",
>           "UMask": "0x1"
>       },
> @@ -855,7 +852,6 @@
>           "MSRIndex": "0x1a6,0x1a7",
>           "MSRValue": "0x10003C07F7",
>           "Offcore": "1",
> -        "PublicDescription": "Counts all data/code/rfo reads
(demand & prefetch) hit in the L3 and the snoop to one of the sibling cores hits the line in M state and the line is forwarded",
>           "SampleAfterValue": "100003",
>           "UMask": "0x1"
>       },
> @@ -866,9 +862,8 @@
>           "EventCode": "0xB7, 0xBB",
>           "EventName":
"OFFCORE_RESPONSE.ALL_READS.LLC_HIT.HIT_OTHER_CORE_NO_FWD",
>           "MSRIndex": "0x1a6,0x1a7",
> -        "MSRValue": "0x04003C07F7",
> +        "MSRValue": "0x4003C07F7",
>           "Offcore": "1",
> -        "PublicDescription": "Counts all data/code/rfo reads
(demand & prefetch) hit in the L3 and the snoops to sibling cores hit in either E/S state and the line is not forwarded",
>           "SampleAfterValue": "100003",
>           "UMask": "0x1"
>       },
> @@ -881,7 +876,6 @@
>           "MSRIndex": "0x1a6,0x1a7",
>           "MSRValue": "0x3F803C8FFF",
>           "Offcore": "1",
> -        "PublicDescription": "Counts all requests hit in the L3", >           "SampleAfterValue": "100003",
>           "UMask": "0x1"
>       },
> @@ -894,7 +888,6 @@
>           "MSRIndex": "0x1a6,0x1a7",
>           "MSRValue": "0x10003C0122",
>           "Offcore": "1",
> -        "PublicDescription": "Counts all demand & prefetch RFOs
hit in the L3 and the snoop to one of the sibling cores hits the line in M state and the line is forwarded",
>           "SampleAfterValue": "100003",
>           "UMask": "0x1"
>       },
> @@ -905,9 +898,8 @@
>           "EventCode": "0xB7, 0xBB",
>           "EventName":
"OFFCORE_RESPONSE.ALL_RFO.LLC_HIT.HIT_OTHER_CORE_NO_FWD",
>           "MSRIndex": "0x1a6,0x1a7",
> -        "MSRValue": "0x04003C0122",
> +        "MSRValue": "0x4003C0122",
>           "Offcore": "1",
> -        "PublicDescription": "Counts all demand & prefetch RFOs
hit in the L3 and the snoops to sibling cores hit in either E/S state and the line is not forwarded",
>           "SampleAfterValue": "100003",
>           "UMask": "0x1"
>       },
> @@ -920,7 +912,6 @@
>           "MSRIndex": "0x1a6,0x1a7",
>           "MSRValue": "0x3F803C0002",
>           "Offcore": "1",
> -        "PublicDescription": "Counts all demand data writes
(RFOs) hit in the L3",
>           "SampleAfterValue": "100003",
>           "UMask": "0x1"
>       },
> @@ -933,7 +924,6 @@
>           "MSRIndex": "0x1a6,0x1a7",
>           "MSRValue": "0x10003C0002",
>           "Offcore": "1",
> -        "PublicDescription": "Counts all demand data writes
(RFOs) hit in the L3 and the snoop to one of the sibling cores hits the line in M state and the line is forwarded",
>           "SampleAfterValue": "100003",
>           "UMask": "0x1"
>       },
> @@ -946,7 +936,6 @@
>           "MSRIndex": "0x1a6,0x1a7",
>           "MSRValue": "0x3F803C0200",
>           "Offcore": "1",
> -        "PublicDescription": "Counts prefetch (that bring data
to LLC only) code reads hit in the L3",
>           "SampleAfterValue": "100003",
>           "UMask": "0x1"
>       },
> @@ -959,7 +948,6 @@
>           "MSRIndex": "0x1a6,0x1a7",
>           "MSRValue": "0x3F803C0100",
>           "Offcore": "1",
> -        "PublicDescription": "Counts all prefetch (that bring
data to LLC only) RFOs hit in the L3",
>           "SampleAfterValue": "100003",
>           "UMask": "0x1"
>       },
> @@ -973,4 +961,4 @@
>           "SampleAfterValue": "100003",
>           "UMask": "0x10"
>       }
> -]
> \ No newline at end of file
> +]
> diff --git
a/tools/perf/pmu-events/arch/x86/broadwellx/floating-point.json b/tools/perf/pmu-events/arch/x86/broadwellx/floating-point.json
> index 9ad37dddb354..93bbc8600321 100644
> --- a/tools/perf/pmu-events/arch/x86/broadwellx/floating-point.json > +++ b/tools/perf/pmu-events/arch/x86/broadwellx/floating-point.json > @@ -5,6 +5,7 @@
>           "CounterHTOff": "0,1,2,3",
>           "EventCode": "0xc7",
>           "EventName": "FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE", > +        "PublicDescription": "Number of SSE/AVX computational
128-bit packed double precision floating-point instructions retired; some instructions will count twice as noted below.  Each count represents 2 computation operations, one for each element.  Applies to SSE* and AVX* packed double precision floating-point
instructions: ADD SUB HADD HSUB SUBADD MUL DIV MIN MAX SQRT DPP FM(N)ADD/SUB.  DPP and FM(N)ADD/SUB instructions count twice as they perform 2 calculations per element. The DAZ and FTZ flags in the MXCSR register need to be set when using these events.",
>           "SampleAfterValue": "2000003",
>           "UMask": "0x4"
>       },
> @@ -14,6 +15,7 @@
>           "CounterHTOff": "0,1,2,3",
>           "EventCode": "0xc7",
>           "EventName": "FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE", > +        "PublicDescription": "Number of SSE/AVX computational
128-bit packed single precision floating-point instructions retired; some instructions will count twice as noted below.  Each count represents 4 computation operations, one for each element.  Applies to SSE* and AVX* packed single precision floating-point
instructions: ADD SUB HADD HSUB SUBADD MUL DIV MIN MAX SQRT RSQRT RCP DPP FM(N)ADD/SUB.  DPP and FM(N)ADD/SUB instructions count twice as they perform 2 calculations per element. The DAZ and FTZ flags in the MXCSR register need to be set when using these events.",
>           "SampleAfterValue": "2000003",
>           "UMask": "0x8"
>       },
> @@ -23,6 +25,7 @@
>           "CounterHTOff": "0,1,2,3",
>           "EventCode": "0xc7",
>           "EventName": "FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE", > +        "PublicDescription": "Number of SSE/AVX computational
256-bit packed double precision floating-point instructions retired; some instructions will count twice as noted below.  Each count represents 4 computation operations, one for each element.  Applies to SSE* and AVX* packed double precision floating-point
instructions: ADD SUB HADD HSUB SUBADD MUL DIV MIN MAX SQRT FM(N)ADD/SUB.  FM(N)ADD/SUB instructions count twice as they perform 2 calculations per element. The DAZ and FTZ flags in the MXCSR register need to be set when using these events.",
>           "SampleAfterValue": "2000003",
>           "UMask": "0x10"
>       },
> @@ -32,6 +35,7 @@
>           "CounterHTOff": "0,1,2,3",
>           "EventCode": "0xc7",
>           "EventName": "FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE", > +        "PublicDescription": "Number of SSE/AVX computational
256-bit packed single precision floating-point instructions retired; some instructions will count twice as noted below.  Each count represents 8 computation operations, one for each element.  Applies to SSE* and AVX* packed single precision floating-point
instructions: ADD SUB HADD HSUB SUBADD MUL DIV MIN MAX SQRT RSQRT RCP DPP FM(N)ADD/SUB.  DPP and FM(N)ADD/SUB instructions count twice as they perform 2 calculations per element. The DAZ and FTZ flags in the MXCSR register need to be set when using these events.",
>           "SampleAfterValue": "2000003",
>           "UMask": "0x20"
>       },
> @@ -59,6 +63,7 @@
>           "CounterHTOff": "0,1,2,3",
>           "EventCode": "0xc7",
>           "EventName": "FP_ARITH_INST_RETIRED.SCALAR",
> +        "PublicDescription": "Number of SSE/AVX computational
scalar single precision and double precision floating-point instructions retired; some instructions will count twice as noted below.  Each count represents 1 computational operation. Applies to SSE* and AVX* scalar single precision floating-point instructions:
ADD SUB MUL DIV MIN MAX SQRT RSQRT RCP FM(N)ADD/SUB.  FM(N)ADD/SUB instructions count twice as they perform 2 calculations per element. The DAZ and FTZ flags in the MXCSR register need to be set when using these events.",
>           "SampleAfterValue": "2000003",
>           "UMask": "0x3"
>       },
> @@ -68,6 +73,7 @@
>           "CounterHTOff": "0,1,2,3",
>           "EventCode": "0xc7",
>           "EventName": "FP_ARITH_INST_RETIRED.SCALAR_DOUBLE", > +        "PublicDescription": "Number of SSE/AVX computational
scalar double precision floating-point instructions retired; some instructions will count twice as noted below.  Each count represents 1 computational operation. Applies to SSE* and AVX* scalar double precision floating-point instructions: ADD SUB MUL DIV MIN MAX SQRT FM(N)ADD/SUB.  FM(N)ADD/SUB instructions count twice as they perform 2 calculations per element. The DAZ and FTZ flags in the MXCSR register need to be set when using these events.",
>           "SampleAfterValue": "2000003",
>           "UMask": "0x1"
>       },
> @@ -77,6 +83,7 @@
>           "CounterHTOff": "0,1,2,3",
>           "EventCode": "0xc7",
>           "EventName": "FP_ARITH_INST_RETIRED.SCALAR_SINGLE", > +        "PublicDescription": "Number of SSE/AVX computational
scalar single precision floating-point instructions retired; some instructions will count twice as noted below.  Each count represents 1 computational operation. Applies to SSE* and AVX* scalar single precision floating-point instructions: ADD SUB MUL DIV MIN MAX SQRT RSQRT RCP FM(N)ADD/SUB.  FM(N)ADD/SUB instructions count twice as they perform 2 calculations per element. The DAZ and FTZ flags in the MXCSR register need to be set when using these events.",
>           "SampleAfterValue": "2000003",
>           "UMask": "0x2"
>       },
> @@ -190,4 +197,4 @@
>           "SampleAfterValue": "2000003",
>           "UMask": "0x3"
>       }
> -]
> \ No newline at end of file
> +]
> diff --git
a/tools/perf/pmu-events/arch/x86/broadwellx/frontend.json b/tools/perf/pmu-events/arch/x86/broadwellx/frontend.json
> index f0bcb945ff76..37ce8034b2ed 100644
> --- a/tools/perf/pmu-events/arch/x86/broadwellx/frontend.json > +++ b/tools/perf/pmu-events/arch/x86/broadwellx/frontend.json > @@ -292,4 +292,4 @@
>           "SampleAfterValue": "2000003",
>           "UMask": "0x1"
>       }
> -]
> \ No newline at end of file
> +]
> diff --git
a/tools/perf/pmu-events/arch/x86/broadwellx/memory.json b/tools/perf/pmu-events/arch/x86/broadwellx/memory.json
> index cce993b197e3..545f61f691b9 100644
> --- a/tools/perf/pmu-events/arch/x86/broadwellx/memory.json > +++ b/tools/perf/pmu-events/arch/x86/broadwellx/memory.json > @@ -247,7 +247,6 @@
>           "MSRIndex": "0x1a6,0x1a7",
>           "MSRValue": "0x3FBFC00244",
>           "Offcore": "1",
> -        "PublicDescription": "Counts all demand & prefetch code
reads miss in the L3",
>           "SampleAfterValue": "100003",
>           "UMask": "0x1"
>       },
> @@ -258,9 +257,8 @@
>           "EventCode": "0xB7, 0xBB",
>           "EventName":
"OFFCORE_RESPONSE.ALL_CODE_RD.LLC_MISS.LOCAL_DRAM",
>           "MSRIndex": "0x1a6,0x1a7",
> -        "MSRValue": "0x0604000244",
> +        "MSRValue": "0x604000244",
>           "Offcore": "1",
> -        "PublicDescription": "Counts all demand & prefetch code
reads miss the L3 and the data is returned from local dram",
>           "SampleAfterValue": "100003",
>           "UMask": "0x1"
>       },
> @@ -273,7 +271,6 @@
>           "MSRIndex": "0x1a6,0x1a7",
>           "MSRValue": "0x3FBFC00091",
>           "Offcore": "1",
> -        "PublicDescription": "Counts all demand & prefetch data
reads miss in the L3",
>           "SampleAfterValue": "100003",
>           "UMask": "0x1"
>       },
> @@ -284,9 +281,8 @@
>           "EventCode": "0xB7, 0xBB",
>           "EventName":
"OFFCORE_RESPONSE.ALL_DATA_RD.LLC_MISS.LOCAL_DRAM",
>           "MSRIndex": "0x1a6,0x1a7",
> -        "MSRValue": "0x0604000091",
> +        "MSRValue": "0x604000091",
>           "Offcore": "1",
> -        "PublicDescription": "Counts all demand & prefetch data
reads miss the L3 and the data is returned from local dram",
>           "SampleAfterValue": "100003",
>           "UMask": "0x1"
>       },
> @@ -297,9 +293,8 @@
>           "EventCode": "0xB7, 0xBB",
>           "EventName":
"OFFCORE_RESPONSE.ALL_DATA_RD.LLC_MISS.REMOTE_DRAM",
>           "MSRIndex": "0x1a6,0x1a7",
> -        "MSRValue": "0x063BC00091",
> +        "MSRValue": "0x63BC00091",
>           "Offcore": "1",
> -        "PublicDescription": "Counts all demand & prefetch data
reads miss the L3 and the data is returned from remote dram",
>           "SampleAfterValue": "100003",
>           "UMask": "0x1"
>       },
> @@ -312,7 +307,6 @@
>           "MSRIndex": "0x1a6,0x1a7",
>           "MSRValue": "0x103FC00091",
>           "Offcore": "1",
> -        "PublicDescription": "Counts all demand & prefetch data
reads miss the L3 and the modified data is transferred from remote cache",
>           "SampleAfterValue": "100003",
>           "UMask": "0x1"
>       },
> @@ -323,9 +317,8 @@
>           "EventCode": "0xB7, 0xBB",
>           "EventName":
"OFFCORE_RESPONSE.ALL_DATA_RD.LLC_MISS.REMOTE_HIT_FORWARD",
>           "MSRIndex": "0x1a6,0x1a7",
> -        "MSRValue": "0x087FC00091",
> +        "MSRValue": "0x87FC00091",
>           "Offcore": "1",
> -        "PublicDescription": "Counts all demand & prefetch data
reads miss the L3 and clean or shared data is transferred from remote cache",
>           "SampleAfterValue": "100003",
>           "UMask": "0x1"
>       },
> @@ -338,20 +331,18 @@
>           "MSRIndex": "0x1a6,0x1a7",
>           "MSRValue": "0x3FBFC007F7",
>           "Offcore": "1",
> -        "PublicDescription": "Counts all data/code/rfo reads
(demand & prefetch) miss in the L3",
>           "SampleAfterValue": "100003",
>           "UMask": "0x1"
>       },
>       {
> -        "BriefDescription": "Counts all data/code/rfo reads
(demand & prefetch)miss the L3 and the data is returned from local dram",
> +        "BriefDescription": "Counts all data/code/rfo reads
(demand & prefetch) miss the L3 and the data is returned from local dram",
>           "Counter": "0,1,2,3",
>           "CounterHTOff": "0,1,2,3",
>           "EventCode": "0xB7, 0xBB",
>           "EventName":
"OFFCORE_RESPONSE.ALL_READS.LLC_MISS.LOCAL_DRAM",
>           "MSRIndex": "0x1a6,0x1a7",
> -        "MSRValue": "0x06040007F7",
> +        "MSRValue": "0x6040007F7",
>           "Offcore": "1",
> -        "PublicDescription": "Counts all data/code/rfo reads
(demand & prefetch)miss the L3 and the data is returned from local dram",
>           "SampleAfterValue": "100003",
>           "UMask": "0x1"
>       },
> @@ -362,9 +353,8 @@
>           "EventCode": "0xB7, 0xBB",
>           "EventName":
"OFFCORE_RESPONSE.ALL_READS.LLC_MISS.REMOTE_DRAM",
>           "MSRIndex": "0x1a6,0x1a7",
> -        "MSRValue": "0x063BC007F7",
> +        "MSRValue": "0x63BC007F7",
>           "Offcore": "1",
> -        "PublicDescription": "Counts all data/code/rfo reads
(demand & prefetch) miss the L3 and the data is returned from remote dram",
>           "SampleAfterValue": "100003",
>           "UMask": "0x1"
>       },
> @@ -377,7 +367,6 @@
>           "MSRIndex": "0x1a6,0x1a7",
>           "MSRValue": "0x103FC007F7",
>           "Offcore": "1",
> -        "PublicDescription": "Counts all data/code/rfo reads
(demand & prefetch) miss the L3 and the modified data is transferred from remote cache",
>           "SampleAfterValue": "100003",
>           "UMask": "0x1"
>       },
> @@ -388,9 +377,8 @@
>           "EventCode": "0xB7, 0xBB",
>           "EventName":
"OFFCORE_RESPONSE.ALL_READS.LLC_MISS.REMOTE_HIT_FORWARD",
>           "MSRIndex": "0x1a6,0x1a7",
> -        "MSRValue": "0x087FC007F7",
> +        "MSRValue": "0x87FC007F7",
>           "Offcore": "1",
> -        "PublicDescription": "Counts all data/code/rfo reads
(demand & prefetch) miss the L3 and clean or shared data is transferred from remote cache",
>           "SampleAfterValue": "100003",
>           "UMask": "0x1"
>       },
> @@ -403,7 +391,6 @@
>           "MSRIndex": "0x1a6,0x1a7",
>           "MSRValue": "0x3FBFC08FFF",
>           "Offcore": "1",
> -        "PublicDescription": "Counts all requests miss in the L3", >           "SampleAfterValue": "100003",
>           "UMask": "0x1"
>       },
> @@ -416,7 +403,6 @@
>           "MSRIndex": "0x1a6,0x1a7",
>           "MSRValue": "0x3FBFC00122",
>           "Offcore": "1",
> -        "PublicDescription": "Counts all demand & prefetch RFOs
miss in the L3",
>           "SampleAfterValue": "100003",
>           "UMask": "0x1"
>       },
> @@ -427,9 +413,8 @@
>           "EventCode": "0xB7, 0xBB",
>           "EventName":
"OFFCORE_RESPONSE.ALL_RFO.LLC_MISS.LOCAL_DRAM",
>           "MSRIndex": "0x1a6,0x1a7",
> -        "MSRValue": "0x0604000122",
> +        "MSRValue": "0x604000122",
>           "Offcore": "1",
> -        "PublicDescription": "Counts all demand & pref

Signature · 1 line
Zhengjun Xing
author reviewer
#6 lore raw

Re: [PATCH v1 02/31] perf vendor events: Update Intel broadwellx

Quoted 37 lines · Xing Zhengjun

HI Arnaldo,

Quoted 20 lines · Ian Rogers

On Sun, Jul 24, 2022, 6:34 PM Xing Zhengjun <[email protected] <mailto:[email protected]>> wrote:

Hi Ian,

On 7/23/2022 6:32 AM, Ian Rogers wrote:
> Use script at:
>
https://github.com/intel/event-converter-for-linux-perf/blob/master/download_and_gen.py
<https://github.com/intel/event-converter-for-linux-perf/blob/master/download_and_gen.py>
>

It is better to add the event JSON file version and TMA version for the future track. For example, the event list is based on broadwellx JSON file v19, the metrics are based on TMA4.4-full.

Quoted 7 lines

Thanks Xing,

I'll add that in v2. I'd skipped it this time as I'd been adding it to the mapfile. I'll also break apart the tremontx to snowridgex change to match yours. I also will rebase to see if that'll fix the git am issue. Apologies in advance to everyone's inboxes :-)

Hi Arnaldo,

Except for Snowridgex, I also post SPR/ADL/HSX/BDX event list last month, Can these be merged in or do I need to update them? Thanks.

Quoted 7 lines

https://lore.kernel.org/all/[email protected]/
https://lore.kernel.org/all/[email protected]/

https://lore.kernel.org/all/[email protected]/
https://lore.kernel.org/all/[email protected]/
https://lore.kernel.org/all/[email protected]/
https://lore.kernel.org/all/[email protected]/

Thanks Zhengjun,

I think those patches are all stale over what is posted here. Particular issues are:
- fixing the files to only contain ascii characters
- updating the version information in mapfile.csv
- for BDX and SPR (HSX isn't posted yet, although I've done some
private testing) adding in the metrics from:
https://github.com/intel/perfmon-metrics/tree/main/BDX/metrics/perf
https://github.com/intel/perfmon-metrics/tree/main/SPR/metrics/perf
- the event data on 01.org was updated. ADL, HSX and SPR were all
updated last Friday prior to my v1 patches.
- I also tested all the patches on the respective architectures,
hence discovering the fake event parsing fix that is in patch 1.

As stated in the cover letter, the goal is to make it so that what is in the Linux tree exactly matches the download_and_gen.py output, however, there were same changes I made that are in this pull request:
https://github.com/intel/event-converter-for-linux-perf/pull/15

Given the size of the new metrics and particularly the previously missed uncore data, it'd be nice to land the "compression" in:
https://lore.kernel.org/lkml/[email protected]/
which prior to this change was reducing the binary size for x86 by 12.5% as well as reducing over a megabytes worth of dirty pages for relocated data.

Thanks,
Ian

Quoted 1,774 lines
Quoted 1,762 lines

Thanks,
Ian

     > to download and generate the latest events and metrics. Manually copy
     > the broadwellx files into perf and update mapfile.csv.
     >
     > Tested with 'perf test':
     >   10: PMU events
         :
     >   10.1: PMU event table sanity
         : Ok
     >   10.2: PMU event map aliases
          : Ok
     >   10.3: Parsing of PMU event table metrics
         : Ok
     >   10.4: Parsing of PMU event table metrics with fake PMUs
          : Ok
     >   90: perf all metricgroups test
         : Ok
     >   91: perf all metrics test
          : Skip
     >   93: perf all PMU test
          : Ok
     >
     > Signed-off-by: Ian Rogers <[email protected]
    <mailto:[email protected]>>
     > ---
     >   .../arch/x86/broadwellx/bdx-metrics.json      |  570 ++-
     >   .../pmu-events/arch/x86/broadwellx/cache.json |   22 +-
     >   .../arch/x86/broadwellx/floating-point.json   |    9 +-
     >   .../arch/x86/broadwellx/frontend.json         |    2 +-
     >   .../arch/x86/broadwellx/memory.json           |   39 +-
     >   .../pmu-events/arch/x86/broadwellx/other.json |    2 +-
     >   .../arch/x86/broadwellx/pipeline.json         |    4 +-
     >   .../arch/x86/broadwellx/uncore-cache.json     | 3788
    ++++++++++++++++-
     >   .../x86/broadwellx/uncore-interconnect.json   | 1438 ++++++-
     >   .../arch/x86/broadwellx/uncore-memory.json    | 2849 ++++++++++++-
     >   .../arch/x86/broadwellx/uncore-other.json     | 3252 ++++++++++++++
     >   .../arch/x86/broadwellx/uncore-power.json     |  437 +-
     >   .../arch/x86/broadwellx/virtual-memory.json   |    2 +-
     >   tools/perf/pmu-events/arch/x86/mapfile.csv    |    2 +-
     >   14 files changed, 12103 insertions(+), 313 deletions(-)
     >   create mode 100644
    tools/perf/pmu-events/arch/x86/broadwellx/uncore-other.json
     >
     > diff --git
    a/tools/perf/pmu-events/arch/x86/broadwellx/bdx-metrics.json
    b/tools/perf/pmu-events/arch/x86/broadwellx/bdx-metrics.json
     > index b055947c0afe..720ee7c9332d 100644
     > --- a/tools/perf/pmu-events/arch/x86/broadwellx/bdx-metrics.json
     > +++ b/tools/perf/pmu-events/arch/x86/broadwellx/bdx-metrics.json
     > @@ -74,12 +74,6 @@
     >           "MetricGroup": "Branches;Fed;FetchBW",
     >           "MetricName": "UpTB"
     >       },
     > -    {
     > -        "BriefDescription": "Cycles Per Instruction (per Logical
    Processor)",
     > -        "MetricExpr": "1 / (INST_RETIRED.ANY /
    CPU_CLK_UNHALTED.THREAD)",
     > -        "MetricGroup": "Pipeline;Mem",
     > -        "MetricName": "CPI"
     > -    },
     >       {
     >           "BriefDescription": "Per-Logical Processor actual
    clocks when the Logical Processor is active.",
     >           "MetricExpr": "CPU_CLK_UNHALTED.THREAD",
     > @@ -130,43 +124,25 @@
     >           "MetricName": "FLOPc_SMT"
     >       },
     >       {
     > -        "BriefDescription": "Actual per-core usage of the
    Floating Point execution units (regardless of the vector width)",
     > +        "BriefDescription": "Actual per-core usage of the
    Floating Point non-X87 execution units (regardless of precision or
    vector-width)",
     >           "MetricExpr": "( (FP_ARITH_INST_RETIRED.SCALAR_SINGLE +
    FP_ARITH_INST_RETIRED.SCALAR_DOUBLE) +
    (FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE +
    FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE +
    FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE +
    FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE) ) / ( 2 *
    CPU_CLK_UNHALTED.THREAD )",
     >           "MetricGroup": "Cor;Flops;HPC",
     >           "MetricName": "FP_Arith_Utilization",
     > -        "PublicDescription": "Actual per-core usage of the
    Floating Point execution units (regardless of the vector width).
    Values > 1 are possible due to Fused-Multiply Add (FMA) counting."
     > +        "PublicDescription": "Actual per-core usage of the
    Floating Point non-X87 execution units (regardless of precision or
    vector-width). Values > 1 are possible due to ([BDW+] Fused-Multiply
    Add (FMA) counting - common; [ADL+] use all of ADD/MUL/FMA in Scalar
    or 128/256-bit vectors - less common)."
     >       },
     >       {
     > -        "BriefDescription": "Actual per-core usage of the
    Floating Point execution units (regardless of the vector width). SMT
    version; use when SMT is enabled and measuring per logical CPU.",
     > +        "BriefDescription": "Actual per-core usage of the
    Floating Point non-X87 execution units (regardless of precision or
    vector-width). SMT version; use when SMT is enabled and measuring
    per logical CPU.",
     >           "MetricExpr": "( (FP_ARITH_INST_RETIRED.SCALAR_SINGLE +
    FP_ARITH_INST_RETIRED.SCALAR_DOUBLE) +
    (FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE +
    FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE +
    FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE +
    FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE) ) / ( 2 * ( (
    CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 +
    CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ) )",
     >           "MetricGroup": "Cor;Flops;HPC_SMT",
     >           "MetricName": "FP_Arith_Utilization_SMT",
     > -        "PublicDescription": "Actual per-core usage of the
    Floating Point execution units (regardless of the vector width).
    Values > 1 are possible due to Fused-Multiply Add (FMA) counting.
    SMT version; use when SMT is enabled and measuring per logical CPU."
     > +        "PublicDescription": "Actual per-core usage of the
    Floating Point non-X87 execution units (regardless of precision or
    vector-width). Values > 1 are possible due to ([BDW+] Fused-Multiply
    Add (FMA) counting - common; [ADL+] use all of ADD/MUL/FMA in Scalar
    or 128/256-bit vectors - less common). SMT version; use when SMT is
    enabled and measuring per logical CPU."
     >       },
     >       {
     > -        "BriefDescription": "Instruction-Level-Parallelism
    (average number of uops executed when there is at least 1 uop
    executed)",
     > +        "BriefDescription": "Instruction-Level-Parallelism
    (average number of uops executed when there is execution) per-core",
     >           "MetricExpr": "UOPS_EXECUTED.THREAD / ((
    cpu@UOPS_EXECUTED.CORE\\,cmask\\=1@ / 2 ) if #SMT_on else
    UOPS_EXECUTED.CYCLES_GE_1_UOP_EXEC)",
     >           "MetricGroup": "Backend;Cor;Pipeline;PortsUtil",
     >           "MetricName": "ILP"
     >       },
     > -    {
     > -        "BriefDescription": "Branch Misprediction Cost: Fraction
    of TMA slots wasted per non-speculative branch misprediction
    (retired JEClear)",
     > -        "MetricExpr": " ( ((BR_MISP_RETIRED.ALL_BRANCHES / (
    BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT )) * ((
    UOPS_ISSUED.ANY - UOPS_RETIRED.RETIRE_SLOTS + 4 *
    INT_MISC.RECOVERY_CYCLES ) / (4 * CPU_CLK_UNHALTED.THREAD))) + (4 *
    IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 *
    CPU_CLK_UNHALTED.THREAD)) * (BR_MISP_RETIRED.ALL_BRANCHES * (12 * (
    BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT + BACLEARS.ANY )
    / CPU_CLK_UNHALTED.THREAD) / ( BR_MISP_RETIRED.ALL_BRANCHES +
    MACHINE_CLEARS.COUNT + BACLEARS.ANY )) / #(4 *
    IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 *
    CPU_CLK_UNHALTED.THREAD)) ) * (4 * CPU_CLK_UNHALTED.THREAD) /
    BR_MISP_RETIRED.ALL_BRANCHES",
     > -        "MetricGroup": "Bad;BrMispredicts",
     > -        "MetricName": "Branch_Misprediction_Cost"
     > -    },
     > -    {
     > -        "BriefDescription": "Branch Misprediction Cost: Fraction
    of TMA slots wasted per non-speculative branch misprediction
    (retired JEClear)",
     > -        "MetricExpr": " ( ((BR_MISP_RETIRED.ALL_BRANCHES / (
    BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT )) * ((
    UOPS_ISSUED.ANY - UOPS_RETIRED.RETIRE_SLOTS + 4 * (
    INT_MISC.RECOVERY_CYCLES_ANY / 2 ) ) / (4 * ( (
    CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 +
    CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK )
    )))) + (4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * (
    ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 +
    CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )))
    * (BR_MISP_RETIRED.ALL_BRANCHES * (12 * (
    BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT + BACLEARS.ANY )
    / CPU_CLK_UNHALTED.THREAD) / ( BR_MISP_RETIRED.ALL_BRANCHES +
    MACHINE_CLEARS.COUNT + BACLEARS.ANY )) / #(4 *
    IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * ( (
    CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 +
    CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )))
    ) * (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 +
    CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))
    / BR_MISP_RETIRED.ALL_BRANCHES",
     > -        "MetricGroup": "Bad;BrMispredicts_SMT",
     > -        "MetricName": "Branch_Misprediction_Cost_SMT"
     > -    },
     > -    {
     > -        "BriefDescription": "Number of Instructions per
    non-speculative Branch Misprediction (JEClear)",
     > -        "MetricExpr": "INST_RETIRED.ANY /
    BR_MISP_RETIRED.ALL_BRANCHES",
     > -        "MetricGroup": "Bad;BadSpec;BrMispredicts",
     > -        "MetricName": "IpMispredict"
     > -    },
     >       {
     >           "BriefDescription": "Core actual clocks when any
    Logical Processor is active on the Physical Core",
     >           "MetricExpr": "( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1
    + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )",
     > @@ -256,6 +232,18 @@
     >           "MetricGroup": "Summary;TmaL1",
     >           "MetricName": "Instructions"
     >       },
     > +    {
     > +        "BriefDescription": "Average number of Uops retired in
    cycles where at least one uop has retired.",
     > +        "MetricExpr": "UOPS_RETIRED.RETIRE_SLOTS /
    cpu@UOPS_RETIRED.RETIRE_SLOTS\\,cmask\\=1@",
     > +        "MetricGroup": "Pipeline;Ret",
     > +        "MetricName": "Retire"
     > +    },
     > +    {
     > +        "BriefDescription": "",
     > +        "MetricExpr": "UOPS_EXECUTED.THREAD /
    cpu@UOPS_EXECUTED.THREAD\\,cmask\\=1@",
     > +        "MetricGroup": "Cor;Pipeline;PortsUtil;SMT",
     > +        "MetricName": "Execute"
     > +    },
     >       {
     >           "BriefDescription": "Fraction of Uops delivered by the
    DSB (aka Decoded ICache; or Uop Cache)",
     >           "MetricExpr": "IDQ.DSB_UOPS / (( IDQ.DSB_UOPS +
    LSD.UOPS + IDQ.MITE_UOPS + IDQ.MS_UOPS ) )",
     > @@ -263,35 +251,34 @@
     >           "MetricName": "DSB_Coverage"
     >       },
     >       {
     > -        "BriefDescription": "Actual Average Latency for L1
    data-cache miss demand load instructions (in core cycles)",
     > -        "MetricExpr": "L1D_PEND_MISS.PENDING / (
    MEM_LOAD_UOPS_RETIRED.L1_MISS + mem_load_uops_retired.hit_lfb )",
     > -        "MetricGroup": "Mem;MemoryBound;MemoryLat",
     > -        "MetricName": "Load_Miss_Real_Latency",
     > -        "PublicDescription": "Actual Average Latency for L1
    data-cache miss demand load instructions (in core cycles). Latency
    may be overestimated for multi-load instructions - e.g. repeat strings."
     > +        "BriefDescription": "Number of Instructions per
    non-speculative Branch Misprediction (JEClear) (lower number means
    higher occurrence rate)",
     > +        "MetricExpr": "INST_RETIRED.ANY /
    BR_MISP_RETIRED.ALL_BRANCHES",
     > +        "MetricGroup": "Bad;BadSpec;BrMispredicts",
     > +        "MetricName": "IpMispredict"
     >       },
     >       {
     > -        "BriefDescription": "Memory-Level-Parallelism (average
    number of L1 miss demand load when there is at least one such miss.
    Per-Logical Processor)",
     > -        "MetricExpr": "L1D_PEND_MISS.PENDING /
    L1D_PEND_MISS.PENDING_CYCLES",
     > -        "MetricGroup": "Mem;MemoryBound;MemoryBW",
     > -        "MetricName": "MLP"
     > +        "BriefDescription": "Branch Misprediction Cost: Fraction
    of TMA slots wasted per non-speculative branch misprediction
    (retired JEClear)",
     > +        "MetricExpr": " ( ((BR_MISP_RETIRED.ALL_BRANCHES / (
    BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT )) * ((
    UOPS_ISSUED.ANY - UOPS_RETIRED.RETIRE_SLOTS + 4 *
    INT_MISC.RECOVERY_CYCLES ) / (4 * CPU_CLK_UNHALTED.THREAD))) + (4 *
    IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 *
    CPU_CLK_UNHALTED.THREAD)) * (BR_MISP_RETIRED.ALL_BRANCHES * (12 * (
    BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT + BACLEARS.ANY )
    / CPU_CLK_UNHALTED.THREAD) / ( BR_MISP_RETIRED.ALL_BRANCHES +
    MACHINE_CLEARS.COUNT + BACLEARS.ANY )) / #(4 *
    IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 *
    CPU_CLK_UNHALTED.THREAD)) ) * (4 * CPU_CLK_UNHALTED.THREAD) /
    BR_MISP_RETIRED.ALL_BRANCHES",
     > +        "MetricGroup": "Bad;BrMispredicts",
     > +        "MetricName": "Branch_Misprediction_Cost"
     >       },
     >       {
     > -        "BriefDescription": "Average data fill bandwidth to the
    L1 data cache [GB / sec]",
     > -        "MetricExpr": "64 * L1D.REPLACEMENT / 1000000000 /
    duration_time",
     > -        "MetricGroup": "Mem;MemoryBW",
     > -        "MetricName": "L1D_Cache_Fill_BW"
     > +        "BriefDescription": "Branch Misprediction Cost: Fraction
    of TMA slots wasted per non-speculative branch misprediction
    (retired JEClear)",
     > +        "MetricExpr": " ( ((BR_MISP_RETIRED.ALL_BRANCHES / (
    BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT )) * ((
    UOPS_ISSUED.ANY - UOPS_RETIRED.RETIRE_SLOTS + 4 * (
    INT_MISC.RECOVERY_CYCLES_ANY / 2 ) ) / (4 * ( (
    CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 +
    CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK )
    )))) + (4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * (
    ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 +
    CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )))
    * (BR_MISP_RETIRED.ALL_BRANCHES * (12 * (
    BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT + BACLEARS.ANY )
    / CPU_CLK_UNHALTED.THREAD) / ( BR_MISP_RETIRED.ALL_BRANCHES +
    MACHINE_CLEARS.COUNT + BACLEARS.ANY )) / #(4 *
    IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * ( (
    CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 +
    CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )))
    ) * (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 +
    CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))
    / BR_MISP_RETIRED.ALL_BRANCHES",
     > +        "MetricGroup": "Bad;BrMispredicts_SMT",
     > +        "MetricName": "Branch_Misprediction_Cost_SMT"
     >       },
     >       {
     > -        "BriefDescription": "Average data fill bandwidth to the
    L2 cache [GB / sec]",
     > -        "MetricExpr": "64 * L2_LINES_IN.ALL / 1000000000 /
    duration_time",
     > -        "MetricGroup": "Mem;MemoryBW",
     > -        "MetricName": "L2_Cache_Fill_BW"
     > +        "BriefDescription": "Actual Average Latency for L1
    data-cache miss demand load operations (in core cycles)",
     > +        "MetricExpr": "L1D_PEND_MISS.PENDING / (
    MEM_LOAD_UOPS_RETIRED.L1_MISS + mem_load_uops_retired.hit_lfb )",
     > +        "MetricGroup": "Mem;MemoryBound;MemoryLat",
     > +        "MetricName": "Load_Miss_Real_Latency"
     >       },
     >       {
     > -        "BriefDescription": "Average per-core data fill
    bandwidth to the L3 cache [GB / sec]",
     > -        "MetricExpr": "64 * LONGEST_LAT_CACHE.MISS / 1000000000
    / duration_time",
     > -        "MetricGroup": "Mem;MemoryBW",
     > -        "MetricName": "L3_Cache_Fill_BW"
     > +        "BriefDescription": "Memory-Level-Parallelism (average
    number of L1 miss demand load when there is at least one such miss.
    Per-Logical Processor)",
     > +        "MetricExpr": "L1D_PEND_MISS.PENDING /
    L1D_PEND_MISS.PENDING_CYCLES",
     > +        "MetricGroup": "Mem;MemoryBound;MemoryBW",
     > +        "MetricName": "MLP"
     >       },
     >       {
     >           "BriefDescription": "L1 cache true misses per kilo
    instruction for retired demand loads",
     > @@ -306,13 +293,13 @@
     >           "MetricName": "L2MPKI"
     >       },
     >       {
     > -        "BriefDescription": "L2 cache misses per kilo
    instruction for all request types (including speculative)",
     > +        "BriefDescription": "L2 cache ([RKL+] true) misses per
    kilo instruction for all request types (including speculative)",
     >           "MetricExpr": "1000 * L2_RQSTS.MISS / INST_RETIRED.ANY",
     >           "MetricGroup": "Mem;CacheMisses;Offcore",
     >           "MetricName": "L2MPKI_All"
     >       },
     >       {
     > -        "BriefDescription": "L2 cache misses per kilo
    instruction for all demand loads  (including speculative)",
     > +        "BriefDescription": "L2 cache ([RKL+] true) misses per
    kilo instruction for all demand loads  (including speculative)",
     >           "MetricExpr": "1000 * L2_RQSTS.DEMAND_DATA_RD_MISS /
    INST_RETIRED.ANY",
     >           "MetricGroup": "Mem;CacheMisses",
     >           "MetricName": "L2MPKI_Load"
     > @@ -348,6 +335,48 @@
     >           "MetricGroup": "Mem;MemoryTLB_SMT",
     >           "MetricName": "Page_Walks_Utilization_SMT"
     >       },
     > +    {
     > +        "BriefDescription": "Average per-core data fill
    bandwidth to the L1 data cache [GB / sec]",
     > +        "MetricExpr": "64 * L1D.REPLACEMENT / 1000000000 /
    duration_time",
     > +        "MetricGroup": "Mem;MemoryBW",
     > +        "MetricName": "L1D_Cache_Fill_BW"
     > +    },
     > +    {
     > +        "BriefDescription": "Average per-core data fill
    bandwidth to the L2 cache [GB / sec]",
     > +        "MetricExpr": "64 * L2_LINES_IN.ALL / 1000000000 /
    duration_time",
     > +        "MetricGroup": "Mem;MemoryBW",
     > +        "MetricName": "L2_Cache_Fill_BW"
     > +    },
     > +    {
     > +        "BriefDescription": "Average per-core data fill
    bandwidth to the L3 cache [GB / sec]",
     > +        "MetricExpr": "64 * LONGEST_LAT_CACHE.MISS / 1000000000
    / duration_time",
     > +        "MetricGroup": "Mem;MemoryBW",
     > +        "MetricName": "L3_Cache_Fill_BW"
     > +    },
     > +    {
     > +        "BriefDescription": "Average per-thread data fill
    bandwidth to the L1 data cache [GB / sec]",
     > +        "MetricExpr": "(64 * L1D.REPLACEMENT / 1000000000 /
    duration_time)",
     > +        "MetricGroup": "Mem;MemoryBW",
     > +        "MetricName": "L1D_Cache_Fill_BW_1T"
     > +    },
     > +    {
     > +        "BriefDescription": "Average per-thread data fill
    bandwidth to the L2 cache [GB / sec]",
     > +        "MetricExpr": "(64 * L2_LINES_IN.ALL / 1000000000 /
    duration_time)",
     > +        "MetricGroup": "Mem;MemoryBW",
     > +        "MetricName": "L2_Cache_Fill_BW_1T"
     > +    },
     > +    {
     > +        "BriefDescription": "Average per-thread data fill
    bandwidth to the L3 cache [GB / sec]",
     > +        "MetricExpr": "(64 * LONGEST_LAT_CACHE.MISS / 1000000000
    / duration_time)",
     > +        "MetricGroup": "Mem;MemoryBW",
     > +        "MetricName": "L3_Cache_Fill_BW_1T"
     > +    },
     > +    {
     > +        "BriefDescription": "Average per-thread data access
    bandwidth to the L3 cache [GB / sec]",
     > +        "MetricExpr": "0",
     > +        "MetricGroup": "Mem;MemoryBW;Offcore",
     > +        "MetricName": "L3_Cache_Access_BW_1T"
     > +    },
     >       {
     >           "BriefDescription": "Average CPU Utilization",
     >           "MetricExpr": "CPU_CLK_UNHALTED.REF_TSC / msr@tsc@",
     > @@ -364,7 +393,8 @@
     >           "BriefDescription": "Giga Floating Point Operations Per
    Second",
     >           "MetricExpr": "( ( 1 * (
    FP_ARITH_INST_RETIRED.SCALAR_SINGLE +
    FP_ARITH_INST_RETIRED.SCALAR_DOUBLE ) + 2 *
    FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + 4 * (
    FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE +
    FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE ) + 8 *
    FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE ) / 1000000000 ) /
    duration_time",
     >           "MetricGroup": "Cor;Flops;HPC",
     > -        "MetricName": "GFLOPs"
     > +        "MetricName": "GFLOPs",
     > +        "PublicDescription": "Giga Floating Point Operations Per
    Second. Aggregate across all supported options of: FP precisions,
    scalar and vector instructions, vector-width and AMX engine."
     >       },
     >       {
     >           "BriefDescription": "Average Frequency Utilization
    relative nominal frequency",
     > @@ -461,5 +491,439 @@
     >           "MetricExpr": "(cstate_pkg@c7\\-residency@ / msr@tsc@)
    * 100",
     >           "MetricGroup": "Power",
     >           "MetricName": "C7_Pkg_Residency"
     > +    },
     > +    {
     > +        "BriefDescription": "CPU operating frequency (in GHz)",
     > +        "MetricExpr": "( CPU_CLK_UNHALTED.THREAD /
    CPU_CLK_UNHALTED.REF_TSC * #SYSTEM_TSC_FREQ ) / 1000000000",
     > +        "MetricGroup": "",
     > +        "MetricName": "cpu_operating_frequency",
     > +        "ScaleUnit": "1GHz"
     > +    },
     > +    {
     > +        "BriefDescription": "Cycles per instruction retired;
    indicating how much time each executed instruction took; in units of
    cycles.",
     > +        "MetricExpr": "CPU_CLK_UNHALTED.THREAD / INST_RETIRED.ANY",
     > +        "MetricGroup": "",
     > +        "MetricName": "cpi",
     > +        "ScaleUnit": "1per_instr"
     > +    },
     > +    {
     > +        "BriefDescription": "The ratio of number of completed
    memory load instructions to the total number completed instructions",
     > +        "MetricExpr": "MEM_UOPS_RETIRED.ALL_LOADS /
    INST_RETIRED.ANY",
     > +        "MetricGroup": "",
     > +        "MetricName": "loads_per_instr",
     > +        "ScaleUnit": "1per_instr"
     > +    },
     > +    {
     > +        "BriefDescription": "The ratio of number of completed
    memory store instructions to the total number completed instructions",
     > +        "MetricExpr": "MEM_UOPS_RETIRED.ALL_STORES /
    INST_RETIRED.ANY",
     > +        "MetricGroup": "",
     > +        "MetricName": "stores_per_instr",
     > +        "ScaleUnit": "1per_instr"
     > +    },
     > +    {
     > +        "BriefDescription": "Ratio of number of requests missing
    L1 data cache (includes data+rfo w/ prefetches) to the total number
    of completed instructions",
     > +        "MetricExpr": "L1D.REPLACEMENT / INST_RETIRED.ANY",
     > +        "MetricGroup": "",
     > +        "MetricName":
    "l1d_mpi_includes_data_plus_rfo_with_prefetches",
     > +        "ScaleUnit": "1per_instr"
     > +    },
     > +    {
     > +        "BriefDescription": "Ratio of number of demand load
    requests hitting in L1 data cache to the total number of completed
    instructions",
     > +        "MetricExpr": "MEM_LOAD_UOPS_RETIRED.L1_HIT /
    INST_RETIRED.ANY",
     > +        "MetricGroup": "",
     > +        "MetricName": "l1d_demand_data_read_hits_per_instr",
     > +        "ScaleUnit": "1per_instr"
     > +    },
     > +    {
     > +        "BriefDescription": "Ratio of number of code read
    requests missing in L1 instruction cache (includes prefetches) to
    the total number of completed instructions",
     > +        "MetricExpr": "L2_RQSTS.ALL_CODE_RD / INST_RETIRED.ANY",
     > +        "MetricGroup": "",
     > +        "MetricName":
    "l1_i_code_read_misses_with_prefetches_per_instr",
     > +        "ScaleUnit": "1per_instr"
     > +    },
     > +    {
     > +        "BriefDescription": "Ratio of number of completed demand
    load requests hitting in L2 cache to the total number of completed
    instructions",
     > +        "MetricExpr": "MEM_LOAD_UOPS_RETIRED.L2_HIT /
    INST_RETIRED.ANY",
     > +        "MetricGroup": "",
     > +        "MetricName": "l2_demand_data_read_hits_per_instr",
     > +        "ScaleUnit": "1per_instr"
     > +    },
     > +    {
     > +        "BriefDescription": "Ratio of number of requests missing
    L2 cache (includes code+data+rfo w/ prefetches) to the total number
    of completed instructions",
     > +        "MetricExpr": "L2_LINES_IN.ALL / INST_RETIRED.ANY",
     > +        "MetricGroup": "",
     > +        "MetricName":
    "l2_mpi_includes_code_plus_data_plus_rfo_with_prefetches",
     > +        "ScaleUnit": "1per_instr"
     > +    },
     > +    {
     > +        "BriefDescription": "Ratio of number of completed data
    read request missing L2 cache to the total number of completed
    instructions",
     > +        "MetricExpr": "MEM_LOAD_UOPS_RETIRED.L2_MISS /
    INST_RETIRED.ANY",
     > +        "MetricGroup": "",
     > +        "MetricName": "l2_demand_data_read_mpi",
     > +        "ScaleUnit": "1per_instr"
     > +    },
     > +    {
     > +        "BriefDescription": "Ratio of number of code read
    request missing L2 cache to the total number of completed instructions",
     > +        "MetricExpr": "L2_RQSTS.CODE_RD_MISS / INST_RETIRED.ANY",
     > +        "MetricGroup": "",
     > +        "MetricName": "l2_demand_code_mpi",
     > +        "ScaleUnit": "1per_instr"
     > +    },
     > +    {
     > +        "BriefDescription": "Ratio of number of data read
    requests missing last level core cache (includes demand w/
    prefetches) to the total number of completed instructions",
     > +        "MetricExpr": "(
    cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x182@ +
    cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x192@ ) /
    INST_RETIRED.ANY",
     > +        "MetricGroup": "",
     > +        "MetricName": "llc_data_read_mpi_demand_plus_prefetch",
     > +        "ScaleUnit": "1per_instr"
     > +    },
     > +    {
     > +        "BriefDescription": "Ratio of number of code read
    requests missing last level core cache (includes demand w/
    prefetches) to the total number of completed instructions",
     > +        "MetricExpr": "(
    cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x181@ +
    cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x191@ ) /
    INST_RETIRED.ANY",
     > +        "MetricGroup": "",
     > +        "MetricName": "llc_code_read_mpi_demand_plus_prefetch",
     > +        "ScaleUnit": "1per_instr"
     > +    },
     > +    {
     > +        "BriefDescription": "Average latency of a last level
    cache (LLC) demand and prefetch data read miss (read memory access)
    in nano seconds",
     > +        "MetricExpr": "( 1000000000 * (
    cbox@UNC_C_TOR_OCCUPANCY.MISS_OPCODE\\,filter_opc\\=0x182@ /
    cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x182@ ) / (
    UNC_C_CLOCKTICKS / ( source_count(UNC_C_CLOCKTICKS) * #num_packages
    ) ) ) * duration_time",
     > +        "MetricGroup": "",
     > +        "MetricName":
    "llc_data_read_demand_plus_prefetch_miss_latency",
     > +        "ScaleUnit": "1ns"
     > +    },
     > +    {
     > +        "BriefDescription": "Average latency of a last level
    cache (LLC) demand and prefetch data read miss (read memory access)
    addressed to local memory in nano seconds",
     > +        "MetricExpr": "( 1000000000 * (
    cbox@UNC_C_TOR_OCCUPANCY.MISS_LOCAL_OPCODE\\,filter_opc\\=0x182@ /
    cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x182@ ) / (
    UNC_C_CLOCKTICKS / ( source_count(UNC_C_CLOCKTICKS) * #num_packages
    ) ) ) * duration_time",
     > +        "MetricGroup": "",
     > +        "MetricName":
    "llc_data_read_demand_plus_prefetch_miss_latency_for_local_requests",
     > +        "ScaleUnit": "1ns"
     > +    },
     > +    {
     > +        "BriefDescription": "Average latency of a last level
    cache (LLC) demand and prefetch data read miss (read memory access)
    addressed to remote memory in nano seconds",
     > +        "MetricExpr": "( 1000000000 * (
    cbox@UNC_C_TOR_OCCUPANCY.MISS_REMOTE_OPCODE\\,filter_opc\\=0x182@ /
    cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x182@ ) / (
    UNC_C_CLOCKTICKS / ( source_count(UNC_C_CLOCKTICKS) * #num_packages
    ) ) ) * duration_time",
     > +        "MetricGroup": "",
     > +        "MetricName":
    "llc_data_read_demand_plus_prefetch_miss_latency_for_remote_requests",
     > +        "ScaleUnit": "1ns"
     > +    },
     > +    {
     > +        "BriefDescription": "Ratio of number of completed page
    walks (for all page sizes) caused by a code fetch to the total
    number of completed instructions. This implies it missed in the ITLB
    (Instruction TLB) and further levels of TLB.",
     > +        "MetricExpr": "ITLB_MISSES.WALK_COMPLETED /
    INST_RETIRED.ANY",
     > +        "MetricGroup": "",
     > +        "MetricName": "itlb_mpi",
     > +        "ScaleUnit": "1per_instr"
     > +    },
     > +    {
     > +        "BriefDescription": "Ratio of number of completed page
    walks (for 2 megabyte and 4 megabyte page sizes) caused by a code
    fetch to the total number of completed instructions. This implies it
    missed in the Instruction Translation Lookaside Buffer (ITLB) and
    further levels of TLB.",
     > +        "MetricExpr": "ITLB_MISSES.WALK_COMPLETED_2M_4M /
    INST_RETIRED.ANY",
     > +        "MetricGroup": "",
     > +        "MetricName": "itlb_large_page_mpi",
     > +        "ScaleUnit": "1per_instr"
     > +    },
     > +    {
     > +        "BriefDescription": "Ratio of number of completed page
    walks (for all page sizes) caused by demand data loads to the total
    number of completed instructions. This implies it missed in the DTLB
    and further levels of TLB.",
     > +        "MetricExpr": "DTLB_LOAD_MISSES.WALK_COMPLETED /
    INST_RETIRED.ANY",
     > +        "MetricGroup": "",
     > +        "MetricName": "dtlb_load_mpi",
     > +        "ScaleUnit": "1per_instr"
     > +    },
     > +    {
     > +        "BriefDescription": "Ratio of number of completed page
    walks (for all page sizes) caused by demand data stores to the total
    number of completed instructions. This implies it missed in the DTLB
    and further levels of TLB.",
     > +        "MetricExpr": "DTLB_STORE_MISSES.WALK_COMPLETED /
    INST_RETIRED.ANY",
     > +        "MetricGroup": "",
     > +        "MetricName": "dtlb_store_mpi",
     > +        "ScaleUnit": "1per_instr"
     > +    },
     > +    {
     > +        "BriefDescription": "Memory read that miss the last
    level cache (LLC) addressed to local DRAM as a percentage of total
    memory read accesses, does not include LLC prefetches.",
     > +        "MetricExpr": "100 *
    cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x182@ / (
    cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x182@ +
    cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x182@ )",
     > +        "MetricGroup": "",
     > +        "MetricName": "numa_percent_reads_addressed_to_local_dram",
     > +        "ScaleUnit": "1%"
     > +    },
     > +    {
     > +        "BriefDescription": "Memory reads that miss the last
    level cache (LLC) addressed to remote DRAM as a percentage of total
    memory read accesses, does not include LLC prefetches.",
     > +        "MetricExpr": "100 *
    cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x182@ / (
    cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x182@ +
    cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x182@ )",
     > +        "MetricGroup": "",
     > +        "MetricName": "numa_percent_reads_addressed_to_remote_dram",
     > +        "ScaleUnit": "1%"
     > +    },
     > +    {
     > +        "BriefDescription": "Uncore operating frequency in GHz",
     > +        "MetricExpr": "UNC_C_CLOCKTICKS / (
    source_count(UNC_C_CLOCKTICKS) * #num_packages ) / 1000000000",
     > +        "MetricGroup": "",
     > +        "MetricName": "uncore_frequency",
     > +        "ScaleUnit": "1GHz"
     > +    },
     > +    {
     > +        "BriefDescription": "Intel(R) Quick Path Interconnect
    (QPI) data transmit bandwidth (MB/sec)",
     > +        "MetricExpr": "( UNC_Q_TxL_FLITS_G0.DATA * 8 / 1000000)
    / duration_time",
     > +        "MetricGroup": "",
     > +        "MetricName": "qpi_data_transmit_bw_only_data",
     > +        "ScaleUnit": "1MB/s"
     > +    },
     > +    {
     > +        "BriefDescription": "DDR memory read bandwidth (MB/sec)",
     > +        "MetricExpr": "( UNC_M_CAS_COUNT.RD * 64 / 1000000) /
    duration_time",
     > +        "MetricGroup": "",
     > +        "MetricName": "memory_bandwidth_read",
     > +        "ScaleUnit": "1MB/s"
     > +    },
     > +    {
     > +        "BriefDescription": "DDR memory write bandwidth (MB/sec)",
     > +        "MetricExpr": "( UNC_M_CAS_COUNT.WR * 64 / 1000000) /
    duration_time",
     > +        "MetricGroup": "",
     > +        "MetricName": "memory_bandwidth_write",
     > +        "ScaleUnit": "1MB/s"
     > +    },
     > +    {
     > +        "BriefDescription": "DDR memory bandwidth (MB/sec)",
     > +        "MetricExpr": "(( UNC_M_CAS_COUNT.RD +
    UNC_M_CAS_COUNT.WR ) * 64 / 1000000) / duration_time",
     > +        "MetricGroup": "",
     > +        "MetricName": "memory_bandwidth_total",
     > +        "ScaleUnit": "1MB/s"
     > +    },
     > +    {
     > +        "BriefDescription": "Bandwidth of IO reads that are
    initiated by end device controllers that are requesting memory from
    the CPU.",
     > +        "MetricExpr": "(
    cbox@UNC_C_TOR_INSERTS.OPCODE\\,filter_opc\\=0x19e@ * 64 / 1000000)
    / duration_time",
     > +        "MetricGroup": "",
     > +        "MetricName": "io_bandwidth_read",
     > +        "ScaleUnit": "1MB/s"
     > +    },
     > +    {
     > +        "BriefDescription": "Bandwidth of IO writes that are
    initiated by end device controllers that are writing memory to the
    CPU.",
     > +        "MetricExpr": "((
    cbox@UNC_C_TOR_INSERTS.OPCODE\\,filter_opc\\=0x1c8\\,filter_tid\\=0x3e@
    +
    cbox@UNC_C_TOR_INSERTS.OPCODE\\,filter_opc\\=0x180\\,filter_tid\\=0x3e@
    ) * 64 / 1000000) / duration_time",
     > +        "MetricGroup": "",
     > +        "MetricName": "io_bandwidth_write",
     > +        "ScaleUnit": "1MB/s"
     > +    },
     > +    {
     > +        "BriefDescription": "Uops delivered from decoded
    instruction cache (decoded stream buffer or DSB) as a percent of
    total uops delivered to Instruction Decode Queue",
     > +        "MetricExpr": "100 * ( IDQ.DSB_UOPS / UOPS_ISSUED.ANY )",
     > +        "MetricGroup": "",
     > +        "MetricName":
    "percent_uops_delivered_from_decoded_icache_dsb",
     > +        "ScaleUnit": "1%"
     > +    },
     > +    {
     > +        "BriefDescription": "Uops delivered from legacy decode
    pipeline (Micro-instruction Translation Engine or MITE) as a percent
    of total uops delivered to Instruction Decode Queue",
     > +        "MetricExpr": "100 * ( IDQ.MITE_UOPS / UOPS_ISSUED.ANY )",
     > +        "MetricGroup": "",
     > +        "MetricName":
    "percent_uops_delivered_from_legacy_decode_pipeline_mite",
     > +        "ScaleUnit": "1%"
     > +    },
     > +    {
     > +        "BriefDescription": "Uops delivered from microcode
    sequencer (MS) as a percent of total uops delivered to Instruction
    Decode Queue",
     > +        "MetricExpr": "100 * ( IDQ.MS_UOPS / UOPS_ISSUED.ANY )",
     > +        "MetricGroup": "",
     > +        "MetricName":
    "percent_uops_delivered_from_microcode_sequencer_ms",
     > +        "ScaleUnit": "1%"
     > +    },
     > +    {
     > +        "BriefDescription": "Uops delivered from loop stream
    detector(LSD) as a percent of total uops delivered to Instruction
    Decode Queue",
     > +        "MetricExpr": "100 * ( LSD.UOPS / UOPS_ISSUED.ANY )",
     > +        "MetricGroup": "",
     > +        "MetricName":
    "percent_uops_delivered_from_loop_stream_detector_lsd",
     > +        "ScaleUnit": "1%"
     > +    },
     > +    {
     > +        "BriefDescription": "This category represents fraction
    of slots where the processor's Frontend undersupplies its Backend.
    Frontend denotes the first part of the processor core responsible to
    fetch operations that are executed later on by the Backend part.
    Within the Frontend; a branch predictor predicts the next address to
    fetch; cache-lines are fetched from the memory subsystem; parsed
    into instructions; and lastly decoded into micro-operations (uops).
    Ideally the Frontend can issue Machine_Width uops every cycle to the
    Backend. Frontend Bound denotes unutilized issue-slots when there is
    no Backend stall; i.e. bubbles where Frontend delivered no uops
    while Backend could have accepted them. For example; stalls due to
    instruction-cache misses would be categorized under Frontend Bound.",
     > +        "MetricExpr": "100 * ( IDQ_UOPS_NOT_DELIVERED.CORE / ( (
    4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else (
    CPU_CLK_UNHALTED.THREAD ) ) ) )",
     > +        "MetricGroup": "TmaL1;PGO",
     > +        "MetricName": "tma_frontend_bound_percent",
     > +        "ScaleUnit": "1%"
     > +    },
     > +    {
     > +        "BriefDescription": "This metric represents fraction of
    slots the CPU was stalled due to Frontend latency issues.  For
    example; instruction-cache misses; iTLB misses or fetch stalls after
    a branch misprediction are categorized under Frontend Latency. In
    such cases; the Frontend eventually delivers no uops for some period.",
     > +        "MetricExpr": "100 * ( ( 4 ) *
    IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / ( ( 4 ) * ( (
    CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else (
    CPU_CLK_UNHALTED.THREAD ) ) ) )",
     > +        "MetricGroup":
    "Frontend;TmaL2;m_tma_frontend_bound_percent",
     > +        "MetricName": "tma_fetch_latency_percent",
     > +        "ScaleUnit": "1%"
     > +    },
     > +    {
     > +        "BriefDescription": "This metric represents fraction of
    cycles the CPU was stalled due to instruction cache misses.",
     > +        "MetricExpr": "100 * ( ICACHE.IFDATA_STALL / (
    CPU_CLK_UNHALTED.THREAD ) )",
     > +        "MetricGroup":
    "BigFoot;FetchLat;IcMiss;TmaL3;m_tma_fetch_latency_percent",
     > +        "MetricName": "tma_icache_misses_percent",
     > +        "ScaleUnit": "1%"
     > +    },
     > +    {
     > +        "BriefDescription": "This metric represents fraction of
    cycles the CPU was stalled due to Instruction TLB (ITLB) misses.",
     > +        "MetricExpr": "100 * ( ( 14 * ITLB_MISSES.STLB_HIT +
    cpu@ITLB_MISSES.WALK_DURATION\\,cmask\\=0x1@ + 7 *
    ITLB_MISSES.WALK_COMPLETED ) / ( CPU_CLK_UNHALTED.THREAD ) )",
     > +        "MetricGroup":
    "BigFoot;FetchLat;MemoryTLB;TmaL3;m_tma_fetch_latency_percent",
     > +        "MetricName": "tma_itlb_misses_percent",
     > +        "ScaleUnit": "1%"
     > +    },
     > +    {
     > +        "BriefDescription": "This metric represents fraction of
    cycles the CPU was stalled due to Branch Resteers. Branch Resteers
    estimates the Frontend delay in fetching operations from corrected
    path; following all sorts of miss-predicted branches. For example;
    branchy code with lots of miss-predictions might get categorized
    under Branch Resteers. Note the value of this node may overlap with
    its siblings.",
     > +        "MetricExpr": "100 * ( ( 12 ) * (
    BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT + BACLEARS.ANY )
    / ( CPU_CLK_UNHALTED.THREAD ) )",
     > +        "MetricGroup": "FetchLat;TmaL3;m_tma_fetch_latency_percent",
     > +        "MetricName": "tma_branch_resteers_percent",
     > +        "ScaleUnit": "1%"
     > +    },
     > +    {
     > +        "BriefDescription": "This metric represents fraction of
    cycles the CPU was stalled due to switches from DSB to MITE
    pipelines. The DSB (decoded i-cache) is a Uop Cache where the
    front-end directly delivers Uops (micro operations) avoiding heavy
    x86 decoding. The DSB pipeline has shorter latency and delivered
    higher bandwidth than the MITE (legacy instruction decode pipeline).
    Switching between the two pipelines can cause penalties hence this
    metric measures the exposed penalty.",
     > +        "MetricExpr": "100 * ( DSB2MITE_SWITCHES.PENALTY_CYCLES
    / ( CPU_CLK_UNHALTED.THREAD ) )",
     > +        "MetricGroup":
    "DSBmiss;FetchLat;TmaL3;m_tma_fetch_latency_percent",
     > +        "MetricName": "tma_dsb_switches_percent",
     > +        "ScaleUnit": "1%"
     > +    },
     > +    {
     > +        "BriefDescription": "This metric represents fraction of
    cycles CPU was stalled due to Length Changing Prefixes (LCPs). Using
    proper compiler flags or Intel Compiler by default will certainly
    avoid this. #Link: Optimization Guide about LCP BKMs.",
     > +        "MetricExpr": "100 * ( ILD_STALL.LCP / (
    CPU_CLK_UNHALTED.THREAD ) )",
     > +        "MetricGroup": "FetchLat;TmaL3;m_tma_fetch_latency_percent",
     > +        "MetricName": "tma_lcp_percent",
     > +        "ScaleUnit": "1%"
     > +    },
     > +    {
     > +        "BriefDescription": "This metric estimates the fraction
    of cycles when the CPU was stalled due to switches of uop delivery
    to the Microcode Sequencer (MS). Commonly used instructions are
    optimized for delivery by the DSB (decoded i-cache) or MITE (legacy
    instruction decode) pipelines. Certain operations cannot be handled
    natively by the execution pipeline; and must be performed by
    microcode (small programs injected into the execution stream).
    Switching to the MS too often can negatively impact performance. The
    MS is designated to deliver long uop flows required by CISC
    instructions like CPUID; or uncommon conditions like Floating Point
    Assists when dealing with Denormals.",
     > +        "MetricExpr": "100 * ( ( 2 ) * IDQ.MS_SWITCHES / (
    CPU_CLK_UNHALTED.THREAD ) )",
     > +        "MetricGroup":
    "FetchLat;MicroSeq;TmaL3;m_tma_fetch_latency_percent",
     > +        "MetricName": "tma_ms_switches_percent",
     > +        "ScaleUnit": "1%"
     > +    },
     > +    {
     > +        "BriefDescription": "This metric represents fraction of
    slots the CPU was stalled due to Frontend bandwidth issues.  For
    example; inefficiencies at the instruction decoders; or restrictions
    for caching in the DSB (decoded uops cache) are categorized under
    Fetch Bandwidth. In such cases; the Frontend typically delivers
    suboptimal amount of uops to the Backend.",
     > +        "MetricExpr": "100 * ( ( IDQ_UOPS_NOT_DELIVERED.CORE / (
    ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else (
    CPU_CLK_UNHALTED.THREAD ) ) ) ) - ( ( 4 ) *
    IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / ( ( 4 ) * ( (
    CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else (
    CPU_CLK_UNHALTED.THREAD ) ) ) ) )",
     > +        "MetricGroup":
    "FetchBW;Frontend;TmaL2;m_tma_frontend_bound_percent",
     > +        "MetricName": "tma_fetch_bandwidth_percent",
     > +        "ScaleUnit": "1%"
     > +    },
     > +    {
     > +        "BriefDescription": "This metric represents Core
    fraction of cycles in which CPU was likely limited due to the MITE
    pipeline (the legacy decode pipeline). This pipeline is used for
    code that was not pre-cached in the DSB or LSD. For example;
    inefficiencies due to asymmetric decoders; use of long immediate or
    LCP can manifest as MITE fetch bandwidth bottleneck.",
     > +        "MetricExpr": "100 * ( ( IDQ.ALL_MITE_CYCLES_ANY_UOPS -
    IDQ.ALL_MITE_CYCLES_4_UOPS ) / ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 )
    if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) / 2 )",
     > +        "MetricGroup":
    "DSBmiss;FetchBW;TmaL3;m_tma_fetch_bandwidth_percent",
     > +        "MetricName": "tma_mite_percent",
     > +        "ScaleUnit": "1%"
     > +    },
     > +    {
     > +        "BriefDescription": "This metric represents Core
    fraction of cycles in which CPU was likely limited due to DSB
    (decoded uop cache) fetch pipeline.  For example; inefficient
    utilization of the DSB cache structure or bank conflict when reading
    from it; are categorized here.",
     > +        "MetricExpr": "100 * ( ( IDQ.ALL_DSB_CYCLES_ANY_UOPS -
    IDQ.ALL_DSB_CYCLES_4_UOPS ) / ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 )
    if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) / 2 )",
     > +        "MetricGroup":
    "DSB;FetchBW;TmaL3;m_tma_fetch_bandwidth_percent",
     > +        "MetricName": "tma_dsb_percent",
     > +        "ScaleUnit": "1%"
     > +    },
     > +    {
     > +        "BriefDescription": "This category represents fraction
    of slots wasted due to incorrect speculations. This include slots
    used to issue uops that do not eventually get retired and slots for
    which the issue-pipeline was blocked due to recovery from earlier
    incorrect speculation. For example; wasted work due to
    miss-predicted branches are categorized under Bad Speculation
    category. Incorrect data speculation followed by Memory Ordering
    Nukes is another example.",
     > +        "MetricExpr": "100 * ( ( UOPS_ISSUED.ANY - (
    UOPS_RETIRED.RETIRE_SLOTS ) + ( 4 ) * ( (
    INT_MISC.RECOVERY_CYCLES_ANY / 2 ) if #SMT_on else
    INT_MISC.RECOVERY_CYCLES ) ) / ( ( 4 ) * ( (
    CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else (
    CPU_CLK_UNHALTED.THREAD ) ) ) )",
     > +        "MetricGroup": "TmaL1",
     > +        "MetricName": "tma_bad_speculation_percent",
     > +        "ScaleUnit": "1%"
     > +    },
     > +    {
     > +        "BriefDescription": "This metric represents fraction of
    slots the CPU has wasted due to Branch Misprediction.  These slots
    are either wasted by uops fetched from an incorrectly speculated
    program path; or stalls when the out-of-order part of the machine
    needs to recover its state from a speculative path.",
     > +        "MetricExpr": "100 * ( ( BR_MISP_RETIRED.ALL_BRANCHES /
    ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT ) ) * ( (
    UOPS_ISSUED.ANY - ( UOPS_RETIRED.RETIRE_SLOTS ) + ( 4 ) * ( (
    INT_MISC.RECOVERY_CYCLES_ANY / 2 ) if #SMT_on else
    INT_MISC.RECOVERY_CYCLES ) ) / ( ( 4 ) * ( (
    CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else (
    CPU_CLK_UNHALTED.THREAD ) ) ) ) )",
     > +        "MetricGroup":
    "BadSpec;BrMispredicts;TmaL2;m_tma_bad_speculation_percent",
     > +        "MetricName": "tma_branch_mispredicts_percent",
     > +        "ScaleUnit": "1%"
     > +    },
     > +    {
     > +        "BriefDescription": "This metric represents fraction of
    slots the CPU has wasted due to Machine Clears.  These slots are
    either wasted by uops fetched prior to the clear; or stalls the
    out-of-order portion of the machine needs to recover its state after
    the clear. For example; this can happen due to memory ordering Nukes
    (e.g. Memory Disambiguation) or Self-Modifying-Code (SMC) nukes.",
     > +        "MetricExpr": "100 * ( ( ( UOPS_ISSUED.ANY - (
    UOPS_RETIRED.RETIRE_SLOTS ) + ( 4 ) * ( (
    INT_MISC.RECOVERY_CYCLES_ANY / 2 ) if #SMT_on else
    INT_MISC.RECOVERY_CYCLES ) ) / ( ( 4 ) * ( (
    CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else (
    CPU_CLK_UNHALTED.THREAD ) ) ) ) - ( ( BR_MISP_RETIRED.ALL_BRANCHES /
    ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT ) ) * ( (
    UOPS_ISSUED.ANY - ( UOPS_RETIRED.RETIRE_SLOTS ) + ( 4 ) * ( (
    INT_MISC.RECOVERY_CYCLES_ANY / 2 ) if #SMT_on else
    INT_MISC.RECOVERY_CYCLES ) ) / ( ( 4 ) * ( (
    CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else (
    CPU_CLK_UNHALTED.THREAD ) ) ) ) ) )",
     > +        "MetricGroup":
    "BadSpec;MachineClears;TmaL2;m_tma_bad_speculation_percent",
     > +        "MetricName": "tma_machine_clears_percent",
     > +        "ScaleUnit": "1%"
     > +    },
     > +    {
     > +        "BriefDescription": "This category represents fraction
    of slots where no uops are being delivered due to a lack of required
    resources for accepting new uops in the Backend. Backend is the
    portion of the processor core where the out-of-order scheduler
    dispatches ready uops into their respective execution units; and
    once completed these uops get retired according to program order.
    For example; stalls due to data-cache misses or stalls due to the
    divider unit being overloaded are both categorized under Backend
    Bound. Backend Bound is further divided into two main categories:
    Memory Bound and Core Bound.",
     > +        "MetricExpr": "100 * ( 1 - ( (
    IDQ_UOPS_NOT_DELIVERED.CORE / ( ( 4 ) * ( (
    CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else (
    CPU_CLK_UNHALTED.THREAD ) ) ) ) + ( ( UOPS_ISSUED.ANY - (
    UOPS_RETIRED.RETIRE_SLOTS ) + ( 4 ) * ( (
    INT_MISC.RECOVERY_CYCLES_ANY / 2 ) if #SMT_on else
    INT_MISC.RECOVERY_CYCLES ) ) / ( ( 4 ) * ( (
    CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else (
    CPU_CLK_UNHALTED.THREAD ) ) ) ) + ( ( UOPS_RETIRED.RETIRE_SLOTS ) /
    ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else (
    CPU_CLK_UNHALTED.THREAD ) ) ) ) ) )",
     > +        "MetricGroup": "TmaL1",
     > +        "MetricName": "tma_backend_bound_percent",
     > +        "ScaleUnit": "1%"
     > +    },
     > +    {
     > +        "BriefDescription": "This metric represents fraction of
    slots the Memory subsystem within the Backend was a bottleneck.
    Memory Bound estimates fraction of slots where pipeline is likely
    stalled due to demand load or store instructions. This accounts
    mainly for (1) non-completed in-flight memory demand loads which
    coincides with execution units starvation; in addition to (2) cases
    where stores could impose backpressure on the pipeline when many of
    them get buffered at the same time (less common out of the two).",
     > +        "MetricExpr": "100 * ( ( ( CYCLE_ACTIVITY.STALLS_MEM_ANY
    + RESOURCE_STALLS.SB <http://RESOURCE_STALLS.SB> ) / ( (
    CYCLE_ACTIVITY.STALLS_TOTAL + UOPS_EXECUTED.CYCLES_GE_1_UOP_EXEC - (
    UOPS_EXECUTED.CYCLES_GE_3_UOPS_EXEC if ( ( INST_RETIRED.ANY / (
    CPU_CLK_UNHALTED.THREAD ) ) > 1.8 ) else
    UOPS_EXECUTED.CYCLES_GE_2_UOPS_EXEC ) - ( RS_EVENTS.EMPTY_CYCLES if
    ( ( ( 4 ) * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / ( ( 4
    ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else (
    CPU_CLK_UNHALTED.THREAD ) ) ) ) > 0.1 ) else 0 ) +
    RESOURCE_STALLS.SB <http://RESOURCE_STALLS.SB> ) ) ) * ( 1 - ( (
    IDQ_UOPS_NOT_DELIVERED.CORE / ( ( 4 ) * ( (
    CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else (
    CPU_CLK_UNHALTED.THREAD ) ) ) ) + ( ( UOPS_ISSUED.ANY - (
    UOPS_RETIRED.RETIRE_SLOTS ) + ( 4 ) * ( (
    INT_MISC.RECOVERY_CYCLES_ANY / 2 ) if #SMT_on else
    INT_MISC.RECOVERY_CYCLES ) ) / ( ( 4 ) * ( (
    CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else (
    CPU_CLK_UNHALTED.THREAD ) ) ) ) + ( ( UOPS_RETIRED.RETIRE_SLOTS ) /
    ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else (
    CPU_CLK_UNHALTED.THREAD ) ) ) ) ) ) )",
     > +        "MetricGroup": "Backend;TmaL2;m_tma_backend_bound_percent",
     > +        "MetricName": "tma_memory_bound_percent",
     > +        "ScaleUnit": "1%"
     > +    },
     > +    {
     > +        "BriefDescription": "This metric estimates how often the
    CPU was stalled without loads missing the L1 data cache.  The L1
    data cache typically has the shortest latency.  However; in certain
    cases like loads blocked on older stores; a load might suffer due to
    high latency even though it is being satisfied by the L1. Another
    example is loads who miss in the TLB. These cases are characterized
    by execution unit stalls; while some non-completed demand load lives
    in the machine without having that demand load missing the L1 cache.",
     > +        "MetricExpr": "100 * ( max( (
    CYCLE_ACTIVITY.STALLS_MEM_ANY - CYCLE_ACTIVITY.STALLS_L1D_MISS ) / (
    CPU_CLK_UNHALTED.THREAD ) , 0 ) )",
     > +        "MetricGroup":
    "CacheMisses;MemoryBound;TmaL3mem;TmaL3;m_tma_memory_bound_percent",
     > +        "MetricName": "tma_l1_bound_percent",
     > +        "ScaleUnit": "1%"
     > +    },
     > +    {
     > +        "BriefDescription": "This metric estimates how often the
    CPU was stalled due to L2 cache accesses by loads.  Avoiding cache
    misses (i.e. L1 misses/L2 hits) can improve the latency and increase
    performance.",
     > +        "MetricExpr": "100 * ( ( CYCLE_ACTIVITY.STALLS_L1D_MISS
    - CYCLE_ACTIVITY.STALLS_L2_MISS ) / ( CPU_CLK_UNHALTED.THREAD ) )",
     > +        "MetricGroup":
    "CacheMisses;MemoryBound;TmaL3mem;TmaL3;m_tma_memory_bound_percent",
     > +        "MetricName": "tma_l2_bound_percent",
     > +        "ScaleUnit": "1%"
     > +    },
     > +    {
     > +        "BriefDescription": "This metric estimates how often the
    CPU was stalled due to loads accesses to L3 cache or contended with
    a sibling Core.  Avoiding cache misses (i.e. L2 misses/L3 hits) can
    improve the latency and increase performance.",
     > +        "MetricExpr": "100 * ( ( MEM_LOAD_UOPS_RETIRED.L3_HIT /
    ( MEM_LOAD_UOPS_RETIRED.L3_HIT + ( 7 ) *
    MEM_LOAD_UOPS_RETIRED.L3_MISS ) ) * CYCLE_ACTIVITY.STALLS_L2_MISS /
    ( CPU_CLK_UNHALTED.THREAD ) )",
     > +        "MetricGroup":
    "CacheMisses;MemoryBound;TmaL3mem;TmaL3;m_tma_memory_bound_percent",
     > +        "MetricName": "tma_l3_bound_percent",
     > +        "ScaleUnit": "1%"
     > +    },
     > +    {
     > +        "BriefDescription": "This metric estimates how often the
    CPU was stalled on accesses to external memory (DRAM) by loads.
    Better caching can improve the latency and increase performance.",
     > +        "MetricExpr": "100 * ( min( ( ( 1 - (
    MEM_LOAD_UOPS_RETIRED.L3_HIT / ( MEM_LOAD_UOPS_RETIRED.L3_HIT + ( 7
    ) * MEM_LOAD_UOPS_RETIRED.L3_MISS ) ) ) *
    CYCLE_ACTIVITY.STALLS_L2_MISS / ( CPU_CLK_UNHALTED.THREAD ) ) , ( 1
    ) ) )",
     > +        "MetricGroup":
    "MemoryBound;TmaL3mem;TmaL3;m_tma_memory_bound_percent",
     > +        "MetricName": "tma_dram_bound_percent",
     > +        "ScaleUnit": "1%"
     > +    },
     > +    {
     > +        "BriefDescription": "This metric estimates how often CPU
    was stalled  due to RFO store memory accesses; RFO store issue a
    read-for-ownership request before the write. Even though store
    accesses do not typically stall out-of-order CPUs; there are few
    cases where stores can lead to actual stalls. This metric will be
    flagged should RFO stores be a bottleneck.",
     > +        "MetricExpr": "100 * ( RESOURCE_STALLS.SB
    <http://RESOURCE_STALLS.SB> / ( CPU_CLK_UNHALTED.THREAD ) )",
     > +        "MetricGroup":
    "MemoryBound;TmaL3mem;TmaL3;m_tma_memory_bound_percent",
     > +        "MetricName": "tma_store_bound_percent",
     > +        "ScaleUnit": "1%"
     > +    },
     > +    {
     > +        "BriefDescription": "This metric represents fraction of
    slots where Core non-memory issues were of a bottleneck.  Shortage
    in hardware compute resources; or dependencies in software's
    instructions are both categorized under Core Bound. Hence it may
    indicate the machine ran out of an out-of-order resource; certain
    execution units are overloaded or dependencies in program's data- or
    instruction-flow are limiting the performance (e.g. FP-chained
    long-latency arithmetic operations).",
     > +        "MetricExpr": "100 * ( ( 1 - ( (
    IDQ_UOPS_NOT_DELIVERED.CORE / ( ( 4 ) * ( (
    CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else (
    CPU_CLK_UNHALTED.THREAD ) ) ) ) + ( ( UOPS_ISSUED.ANY - (
    UOPS_RETIRED.RETIRE_SLOTS ) + ( 4 ) * ( (
    INT_MISC.RECOVERY_CYCLES_ANY / 2 ) if #SMT_on else
    INT_MISC.RECOVERY_CYCLES ) ) / ( ( 4 ) * ( (
    CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else (
    CPU_CLK_UNHALTED.THREAD ) ) ) ) + ( ( UOPS_RETIRED.RETIRE_SLOTS ) /
    ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else (
    CPU_CLK_UNHALTED.THREAD ) ) ) ) ) ) - ( ( (
    CYCLE_ACTIVITY.STALLS_MEM_ANY + RESOURCE_STALLS.SB
    <http://RESOURCE_STALLS.SB> ) / ( ( CYCLE_ACTIVITY.STALLS_TOTAL +
    UOPS_EXECUTED.CYCLES_GE_1_UOP_EXEC - (
    UOPS_EXECUTED.CYCLES_GE_3_UOPS_EXEC if ( ( INST_RETIRED.ANY / (
    CPU_CLK_UNHALTED.THREAD ) ) > 1.8 ) else
    UOPS_EXECUTED.CYCLES_GE_2_UOPS_EXEC ) - ( RS_EVENTS.EMPTY_CYCLES if
    ( ( ( 4 ) * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / ( ( 4
    ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else (
    CPU_CLK_UNHALTED.THREAD ) ) ) ) > 0.1 ) else 0 ) +
    RESOURCE_STALLS.SB <http://RESOURCE_STALLS.SB> ) ) ) * ( 1 - ( (
    IDQ_UOPS_NOT_DELIVERED.CORE / ( ( 4 ) * ( (
    CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else (
    CPU_CLK_UNHALTED.THREAD ) ) ) ) + ( ( UOPS_ISSUED.ANY - (
    UOPS_RETIRED.RETIRE_SLOTS ) + ( 4 ) * ( (
    INT_MISC.RECOVERY_CYCLES_ANY / 2 ) if #SMT_on else
    INT_MISC.RECOVERY_CYCLES ) ) / ( ( 4 ) * ( (
    CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else (
    CPU_CLK_UNHALTED.THREAD ) ) ) ) + ( ( UOPS_RETIRED.RETIRE_SLOTS ) /
    ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else (
    CPU_CLK_UNHALTED.THREAD ) ) ) ) ) ) ) )",
     > +        "MetricGroup":
    "Backend;TmaL2;Compute;m_tma_backend_bound_percent",
     > +        "MetricName": "tma_core_bound_percent",
     > +        "ScaleUnit": "1%"
     > +    },
     > +    {
     > +        "BriefDescription": "This metric represents fraction of
    cycles where the Divider unit was active. Divide and square root
    instructions are performed by the Divider unit and can take
    considerably longer latency than integer or Floating Point addition;
    subtraction; or multiplication.",
     > +        "MetricExpr": "100 * ( ARITH.FPU_DIV_ACTIVE / ( (
    CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else (
    CPU_CLK_UNHALTED.THREAD ) ) )",
     > +        "MetricGroup": "TmaL3;m_tma_core_bound_percent",
     > +        "MetricName": "tma_divider_percent",
     > +        "ScaleUnit": "1%"
     > +    },
     > +    {
     > +        "BriefDescription": "This metric estimates fraction of
    cycles the CPU performance was potentially limited due to Core
    computation issues (non divider-related).  Two distinct categories
    can be attributed into this metric: (1) heavy data-dependency among
    contiguous instructions would manifest in this metric - such cases
    are often referred to as low Instruction Level Parallelism (ILP).
    (2) Contention on some hardware execution unit other than Divider.
    For example; when there are too many multiply operations.",
     > +        "MetricExpr": "100 * ( ( ( ( CYCLE_ACTIVITY.STALLS_TOTAL
    + UOPS_EXECUTED.CYCLES_GE_1_UOP_EXEC - (
    UOPS_EXECUTED.CYCLES_GE_3_UOPS_EXEC if ( ( INST_RETIRED.ANY / (
    CPU_CLK_UNHALTED.THREAD ) ) > 1.8 ) else
    UOPS_EXECUTED.CYCLES_GE_2_UOPS_EXEC ) - ( RS_EVENTS.EMPTY_CYCLES if
    ( ( ( 4 ) * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / ( ( 4
    ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else (
    CPU_CLK_UNHALTED.THREAD ) ) ) ) > 0.1 ) else 0 ) +
    RESOURCE_STALLS.SB <http://RESOURCE_STALLS.SB> ) ) -
    RESOURCE_STALLS.SB <http://RESOURCE_STALLS.SB> -
    CYCLE_ACTIVITY.STALLS_MEM_ANY ) / ( CPU_CLK_UNHALTED.THREAD ) )",
     > +        "MetricGroup": "PortsUtil;TmaL3;m_tma_core_bound_percent",
     > +        "MetricName": "tma_ports_utilization_percent",
     > +        "ScaleUnit": "1%"
     > +    },
     > +    {
     > +        "BriefDescription": "This category represents fraction
    of slots utilized by useful work i.e. issued uops that eventually
    get retired. Ideally; all pipeline slots would be attributed to the
    Retiring category.  Retiring of 100% would indicate the maximum
    Pipeline_Width throughput was achieved.  Maximizing Retiring
    typically increases the Instructions-per-cycle (see IPC metric).
    Note that a high Retiring value does not necessary mean there is no
    room for more performance.  For example; Heavy-operations or
    Microcode Assists are categorized under Retiring. They often
    indicate suboptimal performance and can often be optimized or
    avoided. ",
     > +        "MetricExpr": "100 * ( ( UOPS_RETIRED.RETIRE_SLOTS ) / (
    ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else (
    CPU_CLK_UNHALTED.THREAD ) ) ) )",
     > +        "MetricGroup": "TmaL1",
     > +        "MetricName": "tma_retiring_percent",
     > +        "ScaleUnit": "1%"
     > +    },
     > +    {
     > +        "BriefDescription": "This metric represents fraction of
    slots where the CPU was retiring light-weight operations --
    instructions that require no more than one uop (micro-operation).
    This correlates with total number of instructions used by the
    program. A uops-per-instruction (see UPI metric) ratio of 1 or less
    should be expected for decently optimized software running on Intel
    Core/Xeon products. While this often indicates efficient X86
    instructions were executed; high value does not necessarily mean
    better performance cannot be achieved.",
     > +        "MetricExpr": "100 * ( ( ( UOPS_RETIRED.RETIRE_SLOTS ) /
    ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else (
    CPU_CLK_UNHALTED.THREAD ) ) ) ) - ( ( ( ( UOPS_RETIRED.RETIRE_SLOTS
    ) / UOPS_ISSUED.ANY ) * IDQ.MS_UOPS / ( ( 4 ) * ( (
    CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else (
    CPU_CLK_UNHALTED.THREAD ) ) ) ) ) )",
     > +        "MetricGroup": "Retire;TmaL2;m_tma_retiring_percent",
     > +        "MetricName": "tma_light_operations_percent",
     > +        "ScaleUnit": "1%"
     > +    },
     > +    {
     > +        "BriefDescription": "This metric represents overall
    arithmetic floating-point (FP) operations fraction the CPU has
    executed (retired). Note this metric's value may exceed its parent
    due to use of \"Uops\" CountDomain and FMA double-counting.",
     > +        "MetricExpr": "100 * ( ( INST_RETIRED.X87 * ( (
    UOPS_RETIRED.RETIRE_SLOTS ) / INST_RETIRED.ANY ) / (
    UOPS_RETIRED.RETIRE_SLOTS ) ) + ( (
    FP_ARITH_INST_RETIRED.SCALAR_SINGLE +
    FP_ARITH_INST_RETIRED.SCALAR_DOUBLE ) / ( UOPS_RETIRED.RETIRE_SLOTS
    ) ) + ( min( ( ( FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE +
    FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE +
    FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE +
    FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE ) / (
    UOPS_RETIRED.RETIRE_SLOTS ) ) , ( 1 ) ) ) )",
     > +        "MetricGroup": "HPC;TmaL3;m_tma_light_operations_percent",
     > +        "MetricName": "tma_fp_arith_percent",
     > +        "ScaleUnit": "1%"
     > +    },
     > +    {
     > +        "BriefDescription": "This metric represents fraction of
    slots where the CPU was retiring heavy-weight operations --
    instructions that require two or more uops or microcoded sequences.
    This highly-correlates with the uop length of these
    instructions/sequences.",
     > +        "MetricExpr": "100 * ( ( ( ( UOPS_RETIRED.RETIRE_SLOTS )
    / UOPS_ISSUED.ANY ) * IDQ.MS_UOPS / ( ( 4 ) * ( (
    CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else (
    CPU_CLK_UNHALTED.THREAD ) ) ) ) )",
     > +        "MetricGroup": "Retire;TmaL2;m_tma_retiring_percent",
     > +        "MetricName": "tma_heavy_operations_percent",
     > +        "ScaleUnit": "1%"
     > +    },
     > +    {
     > +        "BriefDescription": "This metric represents fraction of
    slots the CPU was retiring uops fetched by the Microcode Sequencer
    (MS) unit.  The MS is used for CISC instructions not supported by
    the default decoders (like repeat move strings; or CPUID); or by
    microcode assists used to address some operation modes (like in
    Floating Point assists). These cases can often be avoided.",
     > +        "MetricExpr": "100 * ( ( ( UOPS_RETIRED.RETIRE_SLOTS ) /
    UOPS_ISSUED.ANY ) * IDQ.MS_UOPS / ( ( 4 ) * ( (
    CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else (
    CPU_CLK_UNHALTED.THREAD ) ) ) )",
     > +        "MetricGroup":
    "MicroSeq;TmaL3;m_tma_heavy_operations_percent",
     > +        "MetricName": "tma_microcode_sequencer_percent",
     > +        "ScaleUnit": "1%"
     >       }
     >   ]
     > diff --git a/tools/perf/pmu-events/arch/x86/broadwellx/cache.json
    b/tools/perf/pmu-events/arch/x86/broadwellx/cache.json
     > index 127abe08362f..2efc4c0ee740 100644
     > --- a/tools/perf/pmu-events/arch/x86/broadwellx/cache.json
     > +++ b/tools/perf/pmu-events/arch/x86/broadwellx/cache.json
     > @@ -814,9 +814,8 @@
     >           "EventCode": "0xB7, 0xBB",
     >           "EventName":
    "OFFCORE_RESPONSE.ALL_CODE_RD.LLC_HIT.HIT_OTHER_CORE_NO_FWD",
     >           "MSRIndex": "0x1a6,0x1a7",
     > -        "MSRValue": "0x04003C0244",
     > +        "MSRValue": "0x4003C0244",
     >           "Offcore": "1",
     > -        "PublicDescription": "Counts all demand & prefetch code
    reads hit in the L3 and the snoops to sibling cores hit in either
    E/S state and the line is not forwarded",
     >           "SampleAfterValue": "100003",
     >           "UMask": "0x1"
     >       },
     > @@ -829,7 +828,6 @@
     >           "MSRIndex": "0x1a6,0x1a7",
     >           "MSRValue": "0x10003C0091",
     >           "Offcore": "1",
     > -        "PublicDescription": "Counts all demand & prefetch data
    reads hit in the L3 and the snoop to one of the sibling cores hits
    the line in M state and the line is forwarded",
     >           "SampleAfterValue": "100003",
     >           "UMask": "0x1"
     >       },
     > @@ -840,9 +838,8 @@
     >           "EventCode": "0xB7, 0xBB",
     >           "EventName":
    "OFFCORE_RESPONSE.ALL_DATA_RD.LLC_HIT.HIT_OTHER_CORE_NO_FWD",
     >           "MSRIndex": "0x1a6,0x1a7",
     > -        "MSRValue": "0x04003C0091",
     > +        "MSRValue": "0x4003C0091",
     >           "Offcore": "1",
     > -        "PublicDescription": "Counts all demand & prefetch data
    reads hit in the L3 and the snoops to sibling cores hit in either
    E/S state and the line is not forwarded",
     >           "SampleAfterValue": "100003",
     >           "UMask": "0x1"
     >       },
     > @@ -855,7 +852,6 @@
     >           "MSRIndex": "0x1a6,0x1a7",
     >           "MSRValue": "0x10003C07F7",
     >           "Offcore": "1",
     > -        "PublicDescription": "Counts all data/code/rfo reads
    (demand & prefetch) hit in the L3 and the snoop to one of the
    sibling cores hits the line in M state and the line is forwarded",
     >           "SampleAfterValue": "100003",
     >           "UMask": "0x1"
     >       },
     > @@ -866,9 +862,8 @@
     >           "EventCode": "0xB7, 0xBB",
     >           "EventName":
    "OFFCORE_RESPONSE.ALL_READS.LLC_HIT.HIT_OTHER_CORE_NO_FWD",
     >           "MSRIndex": "0x1a6,0x1a7",
     > -        "MSRValue": "0x04003C07F7",
     > +        "MSRValue": "0x4003C07F7",
     >           "Offcore": "1",
     > -        "PublicDescription": "Counts all data/code/rfo reads
    (demand & prefetch) hit in the L3 and the snoops to sibling cores
    hit in either E/S state and the line is not forwarded",
     >           "SampleAfterValue": "100003",
     >           "UMask": "0x1"
     >       },
     > @@ -881,7 +876,6 @@
     >           "MSRIndex": "0x1a6,0x1a7",
     >           "MSRValue": "0x3F803C8FFF",
     >           "Offcore": "1",
     > -        "PublicDescription": "Counts all requests hit in the L3",
     >           "SampleAfterValue": "100003",
     >           "UMask": "0x1"
     >       },
     > @@ -894,7 +888,6 @@
     >           "MSRIndex": "0x1a6,0x1a7",
     >           "MSRValue": "0x10003C0122",
     >           "Offcore": "1",
     > -        "PublicDescription": "Counts all demand & prefetch RFOs
    hit in the L3 and the snoop to one of the sibling cores hits the
    line in M state and the line is forwarded",
     >           "SampleAfterValue": "100003",
     >           "UMask": "0x1"
     >       },
     > @@ -905,9 +898,8 @@
     >           "EventCode": "0xB7, 0xBB",
     >           "EventName":
    "OFFCORE_RESPONSE.ALL_RFO.LLC_HIT.HIT_OTHER_CORE_NO_FWD",
     >           "MSRIndex": "0x1a6,0x1a7",
     > -        "MSRValue": "0x04003C0122",
     > +        "MSRValue": "0x4003C0122",
     >           "Offcore": "1",
     > -        "PublicDescription": "Counts all demand & prefetch RFOs
    hit in the L3 and the snoops to sibling cores hit in either E/S
    state and the line is not forwarded",
     >           "SampleAfterValue": "100003",
     >           "UMask": "0x1"
     >       },
     > @@ -920,7 +912,6 @@
     >           "MSRIndex": "0x1a6,0x1a7",
     >           "MSRValue": "0x3F803C0002",
     >           "Offcore": "1",
     > -        "PublicDescription": "Counts all demand data writes
    (RFOs) hit in the L3",
     >           "SampleAfterValue": "100003",
     >           "UMask": "0x1"
     >       },
     > @@ -933,7 +924,6 @@
     >           "MSRIndex": "0x1a6,0x1a7",
     >           "MSRValue": "0x10003C0002",
     >           "Offcore": "1",
     > -        "PublicDescription": "Counts all demand data writes
    (RFOs) hit in the L3 and the snoop to one of the sibling cores hits
    the line in M state and the line is forwarded",
     >           "SampleAfterValue": "100003",
     >           "UMask": "0x1"
     >       },
     > @@ -946,7 +936,6 @@
     >           "MSRIndex": "0x1a6,0x1a7",
     >           "MSRValue": "0x3F803C0200",
     >           "Offcore": "1",
     > -        "PublicDescription": "Counts prefetch (that bring data
    to LLC only) code reads hit in the L3",
     >           "SampleAfterValue": "100003",
     >           "UMask": "0x1"
     >       },
     > @@ -959,7 +948,6 @@
     >           "MSRIndex": "0x1a6,0x1a7",
     >           "MSRValue": "0x3F803C0100",
     >           "Offcore": "1",
     > -        "PublicDescription": "Counts all prefetch (that bring
    data to LLC only) RFOs hit in the L3",
     >           "SampleAfterValue": "100003",
     >           "UMask": "0x1"
     >       },
     > @@ -973,4 +961,4 @@
     >           "SampleAfterValue": "100003",
     >           "UMask": "0x10"
     >       }
     > -]
     > \ No newline at end of file
     > +]
     > diff --git
    a/tools/perf/pmu-events/arch/x86/broadwellx/floating-point.json
    b/tools/perf/pmu-events/arch/x86/broadwellx/floating-point.json
     > index 9ad37dddb354..93bbc8600321 100644
     > --- a/tools/perf/pmu-events/arch/x86/broadwellx/floating-point.json
     > +++ b/tools/perf/pmu-events/arch/x86/broadwellx/floating-point.json
     > @@ -5,6 +5,7 @@
     >           "CounterHTOff": "0,1,2,3",
     >           "EventCode": "0xc7",
     >           "EventName": "FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE",
     > +        "PublicDescription": "Number of SSE/AVX computational
    128-bit packed double precision floating-point instructions retired;
    some instructions will count twice as noted below.  Each count
    represents 2 computation operations, one for each element.  Applies
    to SSE* and AVX* packed double precision floating-point
    instructions: ADD SUB HADD HSUB SUBADD MUL DIV MIN MAX SQRT DPP
    FM(N)ADD/SUB.  DPP and FM(N)ADD/SUB instructions count twice as they
    perform 2 calculations per element. The DAZ and FTZ flags in the
    MXCSR register need to be set when using these events.",
     >           "SampleAfterValue": "2000003",
     >           "UMask": "0x4"
     >       },
     > @@ -14,6 +15,7 @@
     >           "CounterHTOff": "0,1,2,3",
     >           "EventCode": "0xc7",
     >           "EventName": "FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE",
     > +        "PublicDescription": "Number of SSE/AVX computational
    128-bit packed single precision floating-point instructions retired;
    some instructions will count twice as noted below.  Each count
    represents 4 computation operations, one for each element.  Applies
    to SSE* and AVX* packed single precision floating-point
    instructions: ADD SUB HADD HSUB SUBADD MUL DIV MIN MAX SQRT RSQRT
    RCP DPP FM(N)ADD/SUB.  DPP and FM(N)ADD/SUB instructions count twice
    as they perform 2 calculations per element. The DAZ and FTZ flags in
    the MXCSR register need to be set when using these events.",
     >           "SampleAfterValue": "2000003",
     >           "UMask": "0x8"
     >       },
     > @@ -23,6 +25,7 @@
     >           "CounterHTOff": "0,1,2,3",
     >           "EventCode": "0xc7",
     >           "EventName": "FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE",
     > +        "PublicDescription": "Number of SSE/AVX computational
    256-bit packed double precision floating-point instructions retired;
    some instructions will count twice as noted below.  Each count
    represents 4 computation operations, one for each element.  Applies
    to SSE* and AVX* packed double precision floating-point
    instructions: ADD SUB HADD HSUB SUBADD MUL DIV MIN MAX SQRT
    FM(N)ADD/SUB.  FM(N)ADD/SUB instructions count twice as they perform
    2 calculations per element. The DAZ and FTZ flags in the MXCSR
    register need to be set when using these events.",
     >           "SampleAfterValue": "2000003",
     >           "UMask": "0x10"
     >       },
     > @@ -32,6 +35,7 @@
     >           "CounterHTOff": "0,1,2,3",
     >           "EventCode": "0xc7",
     >           "EventName": "FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE",
     > +        "PublicDescription": "Number of SSE/AVX computational
    256-bit packed single precision floating-point instructions retired;
    some instructions will count twice as noted below.  Each count
    represents 8 computation operations, one for each element.  Applies
    to SSE* and AVX* packed single precision floating-point
    instructions: ADD SUB HADD HSUB SUBADD MUL DIV MIN MAX SQRT RSQRT
    RCP DPP FM(N)ADD/SUB.  DPP and FM(N)ADD/SUB instructions count twice
    as they perform 2 calculations per element. The DAZ and FTZ flags in
    the MXCSR register need to be set when using these events.",
     >           "SampleAfterValue": "2000003",
     >           "UMask": "0x20"
     >       },
     > @@ -59,6 +63,7 @@
     >           "CounterHTOff": "0,1,2,3",
     >           "EventCode": "0xc7",
     >           "EventName": "FP_ARITH_INST_RETIRED.SCALAR",
     > +        "PublicDescription": "Number of SSE/AVX computational
    scalar single precision and double precision floating-point
    instructions retired; some instructions will count twice as noted
    below.  Each count represents 1 computational operation. Applies to
    SSE* and AVX* scalar single precision floating-point instructions:
    ADD SUB MUL DIV MIN MAX SQRT RSQRT RCP FM(N)ADD/SUB.  FM(N)ADD/SUB
    instructions count twice as they perform 2 calculations per element.
    The DAZ and FTZ flags in the MXCSR register need to be set when
    using these events.",
     >           "SampleAfterValue": "2000003",
     >           "UMask": "0x3"
     >       },
     > @@ -68,6 +73,7 @@
     >           "CounterHTOff": "0,1,2,3",
     >           "EventCode": "0xc7",
     >           "EventName": "FP_ARITH_INST_RETIRED.SCALAR_DOUBLE",
     > +        "PublicDescription": "Number of SSE/AVX computational
    scalar double precision floating-point instructions retired; some
    instructions will count twice as noted below.  Each count represents
    1 computational operation. Applies to SSE* and AVX* scalar double
    precision floating-point instructions: ADD SUB MUL DIV MIN MAX SQRT
    FM(N)ADD/SUB.  FM(N)ADD/SUB instructions count twice as they perform
    2 calculations per element. The DAZ and FTZ flags in the MXCSR
    register need to be set when using these events.",
     >           "SampleAfterValue": "2000003",
     >           "UMask": "0x1"
     >       },
     > @@ -77,6 +83,7 @@
     >           "CounterHTOff": "0,1,2,3",
     >           "EventCode": "0xc7",
     >           "EventName": "FP_ARITH_INST_RETIRED.SCALAR_SINGLE",
     > +        "PublicDescription": "Number of SSE/AVX computational
    scalar single precision floating-point instructions retired; some
    instructions will count twice as noted below.  Each count represents
    1 computational operation. Applies to SSE* and AVX* scalar single
    precision floating-point instructions: ADD SUB MUL DIV MIN MAX SQRT
    RSQRT RCP FM(N)ADD/SUB.  FM(N)ADD/SUB instructions count twice as
    they perform 2 calculations per element. The DAZ and FTZ flags in
    the MXCSR register need to be set when using these events.",
     >           "SampleAfterValue": "2000003",
     >           "UMask": "0x2"
     >       },
     > @@ -190,4 +197,4 @@
     >           "SampleAfterValue": "2000003",
     >           "UMask": "0x3"
     >       }
     > -]
     > \ No newline at end of file
     > +]
     > diff --git
    a/tools/perf/pmu-events/arch/x86/broadwellx/frontend.json
    b/tools/perf/pmu-events/arch/x86/broadwellx/frontend.json
     > index f0bcb945ff76..37ce8034b2ed 100644
     > --- a/tools/perf/pmu-events/arch/x86/broadwellx/frontend.json
     > +++ b/tools/perf/pmu-events/arch/x86/broadwellx/frontend.json
     > @@ -292,4 +292,4 @@
     >           "SampleAfterValue": "2000003",
     >           "UMask": "0x1"
     >       }
     > -]
     > \ No newline at end of file
     > +]
     > diff --git
    a/tools/perf/pmu-events/arch/x86/broadwellx/memory.json
    b/tools/perf/pmu-events/arch/x86/broadwellx/memory.json
     > index cce993b197e3..545f61f691b9 100644
     > --- a/tools/perf/pmu-events/arch/x86/broadwellx/memory.json
     > +++ b/tools/perf/pmu-events/arch/x86/broadwellx/memory.json
     > @@ -247,7 +247,6 @@
     >           "MSRIndex": "0x1a6,0x1a7",
     >           "MSRValue": "0x3FBFC00244",
     >           "Offcore": "1",
     > -        "PublicDescription": "Counts all demand & prefetch code
    reads miss in the L3",
     >           "SampleAfterValue": "100003",
     >           "UMask": "0x1"
     >       },
     > @@ -258,9 +257,8 @@
     >           "EventCode": "0xB7, 0xBB",
     >           "EventName":
    "OFFCORE_RESPONSE.ALL_CODE_RD.LLC_MISS.LOCAL_DRAM",
     >           "MSRIndex": "0x1a6,0x1a7",
     > -        "MSRValue": "0x0604000244",
     > +        "MSRValue": "0x604000244",
     >           "Offcore": "1",
     > -        "PublicDescription": "Counts all demand & prefetch code
    reads miss the L3 and the data is returned from local dram",
     >           "SampleAfterValue": "100003",
     >           "UMask": "0x1"
     >       },
     > @@ -273,7 +271,6 @@
     >           "MSRIndex": "0x1a6,0x1a7",
     >           "MSRValue": "0x3FBFC00091",
     >           "Offcore": "1",
     > -        "PublicDescription": "Counts all demand & prefetch data
    reads miss in the L3",
     >           "SampleAfterValue": "100003",
     >           "UMask": "0x1"
     >       },
     > @@ -284,9 +281,8 @@
     >           "EventCode": "0xB7, 0xBB",
     >           "EventName":
    "OFFCORE_RESPONSE.ALL_DATA_RD.LLC_MISS.LOCAL_DRAM",
     >           "MSRIndex": "0x1a6,0x1a7",
     > -        "MSRValue": "0x0604000091",
     > +        "MSRValue": "0x604000091",
     >           "Offcore": "1",
     > -        "PublicDescription": "Counts all demand & prefetch data
    reads miss the L3 and the data is returned from local dram",
     >           "SampleAfterValue": "100003",
     >           "UMask": "0x1"
     >       },
     > @@ -297,9 +293,8 @@
     >           "EventCode": "0xB7, 0xBB",
     >           "EventName":
    "OFFCORE_RESPONSE.ALL_DATA_RD.LLC_MISS.REMOTE_DRAM",
     >           "MSRIndex": "0x1a6,0x1a7",
     > -        "MSRValue": "0x063BC00091",
     > +        "MSRValue": "0x63BC00091",
     >           "Offcore": "1",
     > -        "PublicDescription": "Counts all demand & prefetch data
    reads miss the L3 and the data is returned from remote dram",
     >           "SampleAfterValue": "100003",
     >           "UMask": "0x1"
     >       },
     > @@ -312,7 +307,6 @@
     >           "MSRIndex": "0x1a6,0x1a7",
     >           "MSRValue": "0x103FC00091",
     >           "Offcore": "1",
     > -        "PublicDescription": "Counts all demand & prefetch data
    reads miss the L3 and the modified data is transferred from remote
    cache",
     >           "SampleAfterValue": "100003",
     >           "UMask": "0x1"
     >       },
     > @@ -323,9 +317,8 @@
     >           "EventCode": "0xB7, 0xBB",
     >           "EventName":
    "OFFCORE_RESPONSE.ALL_DATA_RD.LLC_MISS.REMOTE_HIT_FORWARD",
     >           "MSRIndex": "0x1a6,0x1a7",
     > -        "MSRValue": "0x087FC00091",
     > +        "MSRValue": "0x87FC00091",
     >           "Offcore": "1",
     > -        "PublicDescription": "Counts all demand & prefetch data
    reads miss the L3 and clean or shared data is transferred from
    remote cache",
     >           "SampleAfterValue": "100003",
     >           "UMask": "0x1"
     >       },
     > @@ -338,20 +331,18 @@
     >           "MSRIndex": "0x1a6,0x1a7",
     >           "MSRValue": "0x3FBFC007F7",
     >           "Offcore": "1",
     > -        "PublicDescription": "Counts all data/code/rfo reads
    (demand & prefetch) miss in the L3",
     >           "SampleAfterValue": "100003",
     >           "UMask": "0x1"
     >       },
     >       {
     > -        "BriefDescription": "Counts all data/code/rfo reads
    (demand & prefetch)miss the L3 and the data is returned from local
    dram",
     > +        "BriefDescription": "Counts all data/code/rfo reads
    (demand & prefetch) miss the L3 and the data is returned from local
    dram",
     >           "Counter": "0,1,2,3",
     >           "CounterHTOff": "0,1,2,3",
     >           "EventCode": "0xB7, 0xBB",
     >           "EventName":
    "OFFCORE_RESPONSE.ALL_READS.LLC_MISS.LOCAL_DRAM",
     >           "MSRIndex": "0x1a6,0x1a7",
     > -        "MSRValue": "0x06040007F7",
     > +        "MSRValue": "0x6040007F7",
     >           "Offcore": "1",
     > -        "PublicDescription": "Counts all data/code/rfo reads
    (demand & prefetch)miss the L3 and the data is returned from local
    dram",
     >           "SampleAfterValue": "100003",
     >           "UMask": "0x1"
     >       },
     > @@ -362,9 +353,8 @@
     >           "EventCode": "0xB7, 0xBB",
     >           "EventName":
    "OFFCORE_RESPONSE.ALL_READS.LLC_MISS.REMOTE_DRAM",
     >           "MSRIndex": "0x1a6,0x1a7",
     > -        "MSRValue": "0x063BC007F7",
     > +        "MSRValue": "0x63BC007F7",
     >           "Offcore": "1",
     > -        "PublicDescription": "Counts all data/code/rfo reads
    (demand & prefetch) miss the L3 and the data is returned from remote
    dram",
     >           "SampleAfterValue": "100003",
     >           "UMask": "0x1"
     >       },
     > @@ -377,7 +367,6 @@
     >           "MSRIndex": "0x1a6,0x1a7",
     >           "MSRValue": "0x103FC007F7",
     >           "Offcore": "1",
     > -        "PublicDescription": "Counts all data/code/rfo reads
    (demand & prefetch) miss the L3 and the modified data is transferred
    from remote cache",
     >           "SampleAfterValue": "100003",
     >           "UMask": "0x1"
     >       },
     > @@ -388,9 +377,8 @@
     >           "EventCode": "0xB7, 0xBB",
     >           "EventName":
    "OFFCORE_RESPONSE.ALL_READS.LLC_MISS.REMOTE_HIT_FORWARD",
     >           "MSRIndex": "0x1a6,0x1a7",
     > -        "MSRValue": "0x087FC007F7",
     > +        "MSRValue": "0x87FC007F7",
     >           "Offcore": "1",
     > -        "PublicDescription": "Counts all data/code/rfo reads
    (demand & prefetch) miss the L3 and clean or shared data is
    transferred from remote cache",
     >           "SampleAfterValue": "100003",
     >           "UMask": "0x1"
     >       },
     > @@ -403,7 +391,6 @@
     >           "MSRIndex": "0x1a6,0x1a7",
     >           "MSRValue": "0x3FBFC08FFF",
     >           "Offcore": "1",
     > -        "PublicDescription": "Counts all requests miss in the L3",
     >           "SampleAfterValue": "100003",
     >           "UMask": "0x1"
     >       },
     > @@ -416,7 +403,6 @@
     >           "MSRIndex": "0x1a6,0x1a7",
     >           "MSRValue": "0x3FBFC00122",
     >           "Offcore": "1",
     > -        "PublicDescription": "Counts all demand & prefetch RFOs
    miss in the L3",
     >           "SampleAfterValue": "100003",
     >           "UMask": "0x1"
     >       },
     > @@ -427,9 +413,8 @@
Quoted 9 lines
     >           "EventCode": "0xB7, 0xBB",
     >           "EventName":
    "OFFCORE_RESPONSE.ALL_RFO.LLC_MISS.LOCAL_DRAM",
     >           "MSRIndex": "0x1a6,0x1a7",
     > -        "MSRValue": "0x0604000122",
     > +        "MSRValue": "0x604000122",
     >           "Offcore": "1",
     > -        "PublicDescription": "Counts all demand & pref

--
Zhengjun Xing

#7 lore raw

Re: [PATCH v1 02/31] perf vendor events: Update Intel broadwellx

Quoted 42 lines · Ian Rogers
Quoted 40 lines · Xing Zhengjun

HI Arnaldo,

Quoted 20 lines · Ian Rogers

On Sun, Jul 24, 2022, 6:34 PM Xing Zhengjun <[email protected] <mailto:[email protected]>> wrote:

Hi Ian,

On 7/23/2022 6:32 AM, Ian Rogers wrote:
> Use script at:
>
https://github.com/intel/event-converter-for-linux-perf/blob/master/download_and_gen.py
<https://github.com/intel/event-converter-for-linux-perf/blob/master/download_and_gen.py>
>

It is better to add the event JSON file version and TMA version for the future track. For example, the event list is based on broadwellx JSON file v19, the metrics are based on TMA4.4-full.

Quoted 7 lines

Thanks Xing,

I'll add that in v2. I'd skipped it this time as I'd been adding it to the mapfile. I'll also break apart the tremontx to snowridgex change to match yours. I also will rebase to see if that'll fix the git am issue. Apologies in advance to everyone's inboxes :-)

Hi Arnaldo,

Except for Snowridgex, I also post SPR/ADL/HSX/BDX event list last month, Can these be merged in or do I need to update them? Thanks.

https://lore.kernel.org/all/[email protected]/
https://lore.kernel.org/all/[email protected]/

Quoted 8 lines
Quoted 4 lines

https://lore.kernel.org/all/[email protected]/
https://lore.kernel.org/all/[email protected]/
https://lore.kernel.org/all/[email protected]/
https://lore.kernel.org/all/[email protected]/

Thanks Zhengjun,

I think those patches are all stale over what is posted here.

I think sometimes the middle version is also useful when we need back trace issues.

Quoted 1,801 lines

Particular issues are:
- fixing the files to only contain ascii characters
- updating the version information in mapfile.csv
- for BDX and SPR (HSX isn't posted yet, although I've done some
private testing) adding in the metrics from:
https://github.com/intel/perfmon-metrics/tree/main/BDX/metrics/perf
https://github.com/intel/perfmon-metrics/tree/main/SPR/metrics/perf
- the event data on 01.org was updated. ADL, HSX and SPR were all
updated last Friday prior to my v1 patches.
- I also tested all the patches on the respective architectures,
hence discovering the fake event parsing fix that is in patch 1.

As stated in the cover letter, the goal is to make it so that what is in the Linux tree exactly matches the download_and_gen.py output, however, there were same changes I made that are in this pull request:
https://github.com/intel/event-converter-for-linux-perf/pull/15

Given the size of the new metrics and particularly the previously missed uncore data, it'd be nice to land the "compression" in:
https://lore.kernel.org/lkml/[email protected]/
which prior to this change was reducing the binary size for x86 by 12.5% as well as reducing over a megabytes worth of dirty pages for relocated data.

Thanks,
Ian

Quoted 1,774 lines
Quoted 1,762 lines

Thanks,
Ian

      > to download and generate the latest events and metrics. Manually copy
      > the broadwellx files into perf and update mapfile.csv.
      >
      > Tested with 'perf test':
      >   10: PMU events
          :
      >   10.1: PMU event table sanity
          : Ok
      >   10.2: PMU event map aliases
           : Ok
      >   10.3: Parsing of PMU event table metrics
          : Ok
      >   10.4: Parsing of PMU event table metrics with fake PMUs
           : Ok
      >   90: perf all metricgroups test
          : Ok
      >   91: perf all metrics test
           : Skip
      >   93: perf all PMU test
           : Ok
      >
      > Signed-off-by: Ian Rogers <[email protected]
     <mailto:[email protected]>>
      > ---
      >   .../arch/x86/broadwellx/bdx-metrics.json      |  570 ++-
      >   .../pmu-events/arch/x86/broadwellx/cache.json |   22 +-
      >   .../arch/x86/broadwellx/floating-point.json   |    9 +-
      >   .../arch/x86/broadwellx/frontend.json         |    2 +-
      >   .../arch/x86/broadwellx/memory.json           |   39 +-
      >   .../pmu-events/arch/x86/broadwellx/other.json |    2 +-
      >   .../arch/x86/broadwellx/pipeline.json         |    4 +-
      >   .../arch/x86/broadwellx/uncore-cache.json     | 3788
     ++++++++++++++++-
      >   .../x86/broadwellx/uncore-interconnect.json   | 1438 ++++++-
      >   .../arch/x86/broadwellx/uncore-memory.json    | 2849 ++++++++++++-
      >   .../arch/x86/broadwellx/uncore-other.json     | 3252 ++++++++++++++
      >   .../arch/x86/broadwellx/uncore-power.json     |  437 +-
      >   .../arch/x86/broadwellx/virtual-memory.json   |    2 +-
      >   tools/perf/pmu-events/arch/x86/mapfile.csv    |    2 +-
      >   14 files changed, 12103 insertions(+), 313 deletions(-)
      >   create mode 100644
     tools/perf/pmu-events/arch/x86/broadwellx/uncore-other.json
      >
      > diff --git
     a/tools/perf/pmu-events/arch/x86/broadwellx/bdx-metrics.json
     b/tools/perf/pmu-events/arch/x86/broadwellx/bdx-metrics.json
      > index b055947c0afe..720ee7c9332d 100644
      > --- a/tools/perf/pmu-events/arch/x86/broadwellx/bdx-metrics.json
      > +++ b/tools/perf/pmu-events/arch/x86/broadwellx/bdx-metrics.json
      > @@ -74,12 +74,6 @@
      >           "MetricGroup": "Branches;Fed;FetchBW",
      >           "MetricName": "UpTB"
      >       },
      > -    {
      > -        "BriefDescription": "Cycles Per Instruction (per Logical
     Processor)",
      > -        "MetricExpr": "1 / (INST_RETIRED.ANY /
     CPU_CLK_UNHALTED.THREAD)",
      > -        "MetricGroup": "Pipeline;Mem",
      > -        "MetricName": "CPI"
      > -    },
      >       {
      >           "BriefDescription": "Per-Logical Processor actual
     clocks when the Logical Processor is active.",
      >           "MetricExpr": "CPU_CLK_UNHALTED.THREAD",
      > @@ -130,43 +124,25 @@
      >           "MetricName": "FLOPc_SMT"
      >       },
      >       {
      > -        "BriefDescription": "Actual per-core usage of the
     Floating Point execution units (regardless of the vector width)",
      > +        "BriefDescription": "Actual per-core usage of the
     Floating Point non-X87 execution units (regardless of precision or
     vector-width)",
      >           "MetricExpr": "( (FP_ARITH_INST_RETIRED.SCALAR_SINGLE +
     FP_ARITH_INST_RETIRED.SCALAR_DOUBLE) +
     (FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE +
     FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE +
     FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE +
     FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE) ) / ( 2 *
     CPU_CLK_UNHALTED.THREAD )",
      >           "MetricGroup": "Cor;Flops;HPC",
      >           "MetricName": "FP_Arith_Utilization",
      > -        "PublicDescription": "Actual per-core usage of the
     Floating Point execution units (regardless of the vector width).
     Values > 1 are possible due to Fused-Multiply Add (FMA) counting."
      > +        "PublicDescription": "Actual per-core usage of the
     Floating Point non-X87 execution units (regardless of precision or
     vector-width). Values > 1 are possible due to ([BDW+] Fused-Multiply
     Add (FMA) counting - common; [ADL+] use all of ADD/MUL/FMA in Scalar
     or 128/256-bit vectors - less common)."
      >       },
      >       {
      > -        "BriefDescription": "Actual per-core usage of the
     Floating Point execution units (regardless of the vector width). SMT
     version; use when SMT is enabled and measuring per logical CPU.",
      > +        "BriefDescription": "Actual per-core usage of the
     Floating Point non-X87 execution units (regardless of precision or
     vector-width). SMT version; use when SMT is enabled and measuring
     per logical CPU.",
      >           "MetricExpr": "( (FP_ARITH_INST_RETIRED.SCALAR_SINGLE +
     FP_ARITH_INST_RETIRED.SCALAR_DOUBLE) +
     (FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE +
     FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE +
     FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE +
     FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE) ) / ( 2 * ( (
     CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 +
     CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ) )",
      >           "MetricGroup": "Cor;Flops;HPC_SMT",
      >           "MetricName": "FP_Arith_Utilization_SMT",
      > -        "PublicDescription": "Actual per-core usage of the
     Floating Point execution units (regardless of the vector width).
     Values > 1 are possible due to Fused-Multiply Add (FMA) counting.
     SMT version; use when SMT is enabled and measuring per logical CPU."
      > +        "PublicDescription": "Actual per-core usage of the
     Floating Point non-X87 execution units (regardless of precision or
     vector-width). Values > 1 are possible due to ([BDW+] Fused-Multiply
     Add (FMA) counting - common; [ADL+] use all of ADD/MUL/FMA in Scalar
     or 128/256-bit vectors - less common). SMT version; use when SMT is
     enabled and measuring per logical CPU."
      >       },
      >       {
      > -        "BriefDescription": "Instruction-Level-Parallelism
     (average number of uops executed when there is at least 1 uop
     executed)",
      > +        "BriefDescription": "Instruction-Level-Parallelism
     (average number of uops executed when there is execution) per-core",
      >           "MetricExpr": "UOPS_EXECUTED.THREAD / ((
     cpu@UOPS_EXECUTED.CORE\\,cmask\\=1@ / 2 ) if #SMT_on else
     UOPS_EXECUTED.CYCLES_GE_1_UOP_EXEC)",
      >           "MetricGroup": "Backend;Cor;Pipeline;PortsUtil",
      >           "MetricName": "ILP"
      >       },
      > -    {
      > -        "BriefDescription": "Branch Misprediction Cost: Fraction
     of TMA slots wasted per non-speculative branch misprediction
     (retired JEClear)",
      > -        "MetricExpr": " ( ((BR_MISP_RETIRED.ALL_BRANCHES / (
     BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT )) * ((
     UOPS_ISSUED.ANY - UOPS_RETIRED.RETIRE_SLOTS + 4 *
     INT_MISC.RECOVERY_CYCLES ) / (4 * CPU_CLK_UNHALTED.THREAD))) + (4 *
     IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 *
     CPU_CLK_UNHALTED.THREAD)) * (BR_MISP_RETIRED.ALL_BRANCHES * (12 * (
     BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT + BACLEARS.ANY )
     / CPU_CLK_UNHALTED.THREAD) / ( BR_MISP_RETIRED.ALL_BRANCHES +
     MACHINE_CLEARS.COUNT + BACLEARS.ANY )) / #(4 *
     IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 *
     CPU_CLK_UNHALTED.THREAD)) ) * (4 * CPU_CLK_UNHALTED.THREAD) /
     BR_MISP_RETIRED.ALL_BRANCHES",
      > -        "MetricGroup": "Bad;BrMispredicts",
      > -        "MetricName": "Branch_Misprediction_Cost"
      > -    },
      > -    {
      > -        "BriefDescription": "Branch Misprediction Cost: Fraction
     of TMA slots wasted per non-speculative branch misprediction
     (retired JEClear)",
      > -        "MetricExpr": " ( ((BR_MISP_RETIRED.ALL_BRANCHES / (
     BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT )) * ((
     UOPS_ISSUED.ANY - UOPS_RETIRED.RETIRE_SLOTS + 4 * (
     INT_MISC.RECOVERY_CYCLES_ANY / 2 ) ) / (4 * ( (
     CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 +
     CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK )
     )))) + (4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * (
     ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 +
     CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )))
     * (BR_MISP_RETIRED.ALL_BRANCHES * (12 * (
     BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT + BACLEARS.ANY )
     / CPU_CLK_UNHALTED.THREAD) / ( BR_MISP_RETIRED.ALL_BRANCHES +
     MACHINE_CLEARS.COUNT + BACLEARS.ANY )) / #(4 *
     IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * ( (
     CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 +
     CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )))
     ) * (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 +
     CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))
     / BR_MISP_RETIRED.ALL_BRANCHES",
      > -        "MetricGroup": "Bad;BrMispredicts_SMT",
      > -        "MetricName": "Branch_Misprediction_Cost_SMT"
      > -    },
      > -    {
      > -        "BriefDescription": "Number of Instructions per
     non-speculative Branch Misprediction (JEClear)",
      > -        "MetricExpr": "INST_RETIRED.ANY /
     BR_MISP_RETIRED.ALL_BRANCHES",
      > -        "MetricGroup": "Bad;BadSpec;BrMispredicts",
      > -        "MetricName": "IpMispredict"
      > -    },
      >       {
      >           "BriefDescription": "Core actual clocks when any
     Logical Processor is active on the Physical Core",
      >           "MetricExpr": "( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1
     + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )",
      > @@ -256,6 +232,18 @@
      >           "MetricGroup": "Summary;TmaL1",
      >           "MetricName": "Instructions"
      >       },
      > +    {
      > +        "BriefDescription": "Average number of Uops retired in
     cycles where at least one uop has retired.",
      > +        "MetricExpr": "UOPS_RETIRED.RETIRE_SLOTS /
     cpu@UOPS_RETIRED.RETIRE_SLOTS\\,cmask\\=1@",
      > +        "MetricGroup": "Pipeline;Ret",
      > +        "MetricName": "Retire"
      > +    },
      > +    {
      > +        "BriefDescription": "",
      > +        "MetricExpr": "UOPS_EXECUTED.THREAD /
     cpu@UOPS_EXECUTED.THREAD\\,cmask\\=1@",
      > +        "MetricGroup": "Cor;Pipeline;PortsUtil;SMT",
      > +        "MetricName": "Execute"
      > +    },
      >       {
      >           "BriefDescription": "Fraction of Uops delivered by the
     DSB (aka Decoded ICache; or Uop Cache)",
      >           "MetricExpr": "IDQ.DSB_UOPS / (( IDQ.DSB_UOPS +
     LSD.UOPS + IDQ.MITE_UOPS + IDQ.MS_UOPS ) )",
      > @@ -263,35 +251,34 @@
      >           "MetricName": "DSB_Coverage"
      >       },
      >       {
      > -        "BriefDescription": "Actual Average Latency for L1
     data-cache miss demand load instructions (in core cycles)",
      > -        "MetricExpr": "L1D_PEND_MISS.PENDING / (
     MEM_LOAD_UOPS_RETIRED.L1_MISS + mem_load_uops_retired.hit_lfb )",
      > -        "MetricGroup": "Mem;MemoryBound;MemoryLat",
      > -        "MetricName": "Load_Miss_Real_Latency",
      > -        "PublicDescription": "Actual Average Latency for L1
     data-cache miss demand load instructions (in core cycles). Latency
     may be overestimated for multi-load instructions - e.g. repeat strings."
      > +        "BriefDescription": "Number of Instructions per
     non-speculative Branch Misprediction (JEClear) (lower number means
     higher occurrence rate)",
      > +        "MetricExpr": "INST_RETIRED.ANY /
     BR_MISP_RETIRED.ALL_BRANCHES",
      > +        "MetricGroup": "Bad;BadSpec;BrMispredicts",
      > +        "MetricName": "IpMispredict"
      >       },
      >       {
      > -        "BriefDescription": "Memory-Level-Parallelism (average
     number of L1 miss demand load when there is at least one such miss.
     Per-Logical Processor)",
      > -        "MetricExpr": "L1D_PEND_MISS.PENDING /
     L1D_PEND_MISS.PENDING_CYCLES",
      > -        "MetricGroup": "Mem;MemoryBound;MemoryBW",
      > -        "MetricName": "MLP"
      > +        "BriefDescription": "Branch Misprediction Cost: Fraction
     of TMA slots wasted per non-speculative branch misprediction
     (retired JEClear)",
      > +        "MetricExpr": " ( ((BR_MISP_RETIRED.ALL_BRANCHES / (
     BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT )) * ((
     UOPS_ISSUED.ANY - UOPS_RETIRED.RETIRE_SLOTS + 4 *
     INT_MISC.RECOVERY_CYCLES ) / (4 * CPU_CLK_UNHALTED.THREAD))) + (4 *
     IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 *
     CPU_CLK_UNHALTED.THREAD)) * (BR_MISP_RETIRED.ALL_BRANCHES * (12 * (
     BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT + BACLEARS.ANY )
     / CPU_CLK_UNHALTED.THREAD) / ( BR_MISP_RETIRED.ALL_BRANCHES +
     MACHINE_CLEARS.COUNT + BACLEARS.ANY )) / #(4 *
     IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 *
     CPU_CLK_UNHALTED.THREAD)) ) * (4 * CPU_CLK_UNHALTED.THREAD) /
     BR_MISP_RETIRED.ALL_BRANCHES",
      > +        "MetricGroup": "Bad;BrMispredicts",
      > +        "MetricName": "Branch_Misprediction_Cost"
      >       },
      >       {
      > -        "BriefDescription": "Average data fill bandwidth to the
     L1 data cache [GB / sec]",
      > -        "MetricExpr": "64 * L1D.REPLACEMENT / 1000000000 /
     duration_time",
      > -        "MetricGroup": "Mem;MemoryBW",
      > -        "MetricName": "L1D_Cache_Fill_BW"
      > +        "BriefDescription": "Branch Misprediction Cost: Fraction
     of TMA slots wasted per non-speculative branch misprediction
     (retired JEClear)",
      > +        "MetricExpr": " ( ((BR_MISP_RETIRED.ALL_BRANCHES / (
     BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT )) * ((
     UOPS_ISSUED.ANY - UOPS_RETIRED.RETIRE_SLOTS + 4 * (
     INT_MISC.RECOVERY_CYCLES_ANY / 2 ) ) / (4 * ( (
     CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 +
     CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK )
     )))) + (4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * (
     ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 +
     CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )))
     * (BR_MISP_RETIRED.ALL_BRANCHES * (12 * (
     BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT + BACLEARS.ANY )
     / CPU_CLK_UNHALTED.THREAD) / ( BR_MISP_RETIRED.ALL_BRANCHES +
     MACHINE_CLEARS.COUNT + BACLEARS.ANY )) / #(4 *
     IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * ( (
     CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 +
     CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )))
     ) * (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 +
     CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))
     / BR_MISP_RETIRED.ALL_BRANCHES",
      > +        "MetricGroup": "Bad;BrMispredicts_SMT",
      > +        "MetricName": "Branch_Misprediction_Cost_SMT"
      >       },
      >       {
      > -        "BriefDescription": "Average data fill bandwidth to the
     L2 cache [GB / sec]",
      > -        "MetricExpr": "64 * L2_LINES_IN.ALL / 1000000000 /
     duration_time",
      > -        "MetricGroup": "Mem;MemoryBW",
      > -        "MetricName": "L2_Cache_Fill_BW"
      > +        "BriefDescription": "Actual Average Latency for L1
     data-cache miss demand load operations (in core cycles)",
      > +        "MetricExpr": "L1D_PEND_MISS.PENDING / (
     MEM_LOAD_UOPS_RETIRED.L1_MISS + mem_load_uops_retired.hit_lfb )",
      > +        "MetricGroup": "Mem;MemoryBound;MemoryLat",
      > +        "MetricName": "Load_Miss_Real_Latency"
      >       },
      >       {
      > -        "BriefDescription": "Average per-core data fill
     bandwidth to the L3 cache [GB / sec]",
      > -        "MetricExpr": "64 * LONGEST_LAT_CACHE.MISS / 1000000000
     / duration_time",
      > -        "MetricGroup": "Mem;MemoryBW",
      > -        "MetricName": "L3_Cache_Fill_BW"
      > +        "BriefDescription": "Memory-Level-Parallelism (average
     number of L1 miss demand load when there is at least one such miss.
     Per-Logical Processor)",
      > +        "MetricExpr": "L1D_PEND_MISS.PENDING /
     L1D_PEND_MISS.PENDING_CYCLES",
      > +        "MetricGroup": "Mem;MemoryBound;MemoryBW",
      > +        "MetricName": "MLP"
      >       },
      >       {
      >           "BriefDescription": "L1 cache true misses per kilo
     instruction for retired demand loads",
      > @@ -306,13 +293,13 @@
      >           "MetricName": "L2MPKI"
      >       },
      >       {
      > -        "BriefDescription": "L2 cache misses per kilo
     instruction for all request types (including speculative)",
      > +        "BriefDescription": "L2 cache ([RKL+] true) misses per
     kilo instruction for all request types (including speculative)",
      >           "MetricExpr": "1000 * L2_RQSTS.MISS / INST_RETIRED.ANY",
      >           "MetricGroup": "Mem;CacheMisses;Offcore",
      >           "MetricName": "L2MPKI_All"
      >       },
      >       {
      > -        "BriefDescription": "L2 cache misses per kilo
     instruction for all demand loads  (including speculative)",
      > +        "BriefDescription": "L2 cache ([RKL+] true) misses per
     kilo instruction for all demand loads  (including speculative)",
      >           "MetricExpr": "1000 * L2_RQSTS.DEMAND_DATA_RD_MISS /
     INST_RETIRED.ANY",
      >           "MetricGroup": "Mem;CacheMisses",
      >           "MetricName": "L2MPKI_Load"
      > @@ -348,6 +335,48 @@
      >           "MetricGroup": "Mem;MemoryTLB_SMT",
      >           "MetricName": "Page_Walks_Utilization_SMT"
      >       },
      > +    {
      > +        "BriefDescription": "Average per-core data fill
     bandwidth to the L1 data cache [GB / sec]",
      > +        "MetricExpr": "64 * L1D.REPLACEMENT / 1000000000 /
     duration_time",
      > +        "MetricGroup": "Mem;MemoryBW",
      > +        "MetricName": "L1D_Cache_Fill_BW"
      > +    },
      > +    {
      > +        "BriefDescription": "Average per-core data fill
     bandwidth to the L2 cache [GB / sec]",
      > +        "MetricExpr": "64 * L2_LINES_IN.ALL / 1000000000 /
     duration_time",
      > +        "MetricGroup": "Mem;MemoryBW",
      > +        "MetricName": "L2_Cache_Fill_BW"
      > +    },
      > +    {
      > +        "BriefDescription": "Average per-core data fill
     bandwidth to the L3 cache [GB / sec]",
      > +        "MetricExpr": "64 * LONGEST_LAT_CACHE.MISS / 1000000000
     / duration_time",
      > +        "MetricGroup": "Mem;MemoryBW",
      > +        "MetricName": "L3_Cache_Fill_BW"
      > +    },
      > +    {
      > +        "BriefDescription": "Average per-thread data fill
     bandwidth to the L1 data cache [GB / sec]",
      > +        "MetricExpr": "(64 * L1D.REPLACEMENT / 1000000000 /
     duration_time)",
      > +        "MetricGroup": "Mem;MemoryBW",
      > +        "MetricName": "L1D_Cache_Fill_BW_1T"
      > +    },
      > +    {
      > +        "BriefDescription": "Average per-thread data fill
     bandwidth to the L2 cache [GB / sec]",
      > +        "MetricExpr": "(64 * L2_LINES_IN.ALL / 1000000000 /
     duration_time)",
      > +        "MetricGroup": "Mem;MemoryBW",
      > +        "MetricName": "L2_Cache_Fill_BW_1T"
      > +    },
      > +    {
      > +        "BriefDescription": "Average per-thread data fill
     bandwidth to the L3 cache [GB / sec]",
      > +        "MetricExpr": "(64 * LONGEST_LAT_CACHE.MISS / 1000000000
     / duration_time)",
      > +        "MetricGroup": "Mem;MemoryBW",
      > +        "MetricName": "L3_Cache_Fill_BW_1T"
      > +    },
      > +    {
      > +        "BriefDescription": "Average per-thread data access
     bandwidth to the L3 cache [GB / sec]",
      > +        "MetricExpr": "0",
      > +        "MetricGroup": "Mem;MemoryBW;Offcore",
      > +        "MetricName": "L3_Cache_Access_BW_1T"
      > +    },
      >       {
      >           "BriefDescription": "Average CPU Utilization",
      >           "MetricExpr": "CPU_CLK_UNHALTED.REF_TSC / msr@tsc@",
      > @@ -364,7 +393,8 @@
      >           "BriefDescription": "Giga Floating Point Operations Per
     Second",
      >           "MetricExpr": "( ( 1 * (
     FP_ARITH_INST_RETIRED.SCALAR_SINGLE +
     FP_ARITH_INST_RETIRED.SCALAR_DOUBLE ) + 2 *
     FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + 4 * (
     FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE +
     FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE ) + 8 *
     FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE ) / 1000000000 ) /
     duration_time",
      >           "MetricGroup": "Cor;Flops;HPC",
      > -        "MetricName": "GFLOPs"
      > +        "MetricName": "GFLOPs",
      > +        "PublicDescription": "Giga Floating Point Operations Per
     Second. Aggregate across all supported options of: FP precisions,
     scalar and vector instructions, vector-width and AMX engine."
      >       },
      >       {
      >           "BriefDescription": "Average Frequency Utilization
     relative nominal frequency",
      > @@ -461,5 +491,439 @@
      >           "MetricExpr": "(cstate_pkg@c7\\-residency@ / msr@tsc@)
     * 100",
      >           "MetricGroup": "Power",
      >           "MetricName": "C7_Pkg_Residency"
      > +    },
      > +    {
      > +        "BriefDescription": "CPU operating frequency (in GHz)",
      > +        "MetricExpr": "( CPU_CLK_UNHALTED.THREAD /
     CPU_CLK_UNHALTED.REF_TSC * #SYSTEM_TSC_FREQ ) / 1000000000",
      > +        "MetricGroup": "",
      > +        "MetricName": "cpu_operating_frequency",
      > +        "ScaleUnit": "1GHz"
      > +    },
      > +    {
      > +        "BriefDescription": "Cycles per instruction retired;
     indicating how much time each executed instruction took; in units of
     cycles.",
      > +        "MetricExpr": "CPU_CLK_UNHALTED.THREAD / INST_RETIRED.ANY",
      > +        "MetricGroup": "",
      > +        "MetricName": "cpi",
      > +        "ScaleUnit": "1per_instr"
      > +    },
      > +    {
      > +        "BriefDescription": "The ratio of number of completed
     memory load instructions to the total number completed instructions",
      > +        "MetricExpr": "MEM_UOPS_RETIRED.ALL_LOADS /
     INST_RETIRED.ANY",
      > +        "MetricGroup": "",
      > +        "MetricName": "loads_per_instr",
      > +        "ScaleUnit": "1per_instr"
      > +    },
      > +    {
      > +        "BriefDescription": "The ratio of number of completed
     memory store instructions to the total number completed instructions",
      > +        "MetricExpr": "MEM_UOPS_RETIRED.ALL_STORES /
     INST_RETIRED.ANY",
      > +        "MetricGroup": "",
      > +        "MetricName": "stores_per_instr",
      > +        "ScaleUnit": "1per_instr"
      > +    },
      > +    {
      > +        "BriefDescription": "Ratio of number of requests missing
     L1 data cache (includes data+rfo w/ prefetches) to the total number
     of completed instructions",
      > +        "MetricExpr": "L1D.REPLACEMENT / INST_RETIRED.ANY",
      > +        "MetricGroup": "",
      > +        "MetricName":
     "l1d_mpi_includes_data_plus_rfo_with_prefetches",
      > +        "ScaleUnit": "1per_instr"
      > +    },
      > +    {
      > +        "BriefDescription": "Ratio of number of demand load
     requests hitting in L1 data cache to the total number of completed
     instructions",
      > +        "MetricExpr": "MEM_LOAD_UOPS_RETIRED.L1_HIT /
     INST_RETIRED.ANY",
      > +        "MetricGroup": "",
      > +        "MetricName": "l1d_demand_data_read_hits_per_instr",
      > +        "ScaleUnit": "1per_instr"
      > +    },
      > +    {
      > +        "BriefDescription": "Ratio of number of code read
     requests missing in L1 instruction cache (includes prefetches) to
     the total number of completed instructions",
      > +        "MetricExpr": "L2_RQSTS.ALL_CODE_RD / INST_RETIRED.ANY",
      > +        "MetricGroup": "",
      > +        "MetricName":
     "l1_i_code_read_misses_with_prefetches_per_instr",
      > +        "ScaleUnit": "1per_instr"
      > +    },
      > +    {
      > +        "BriefDescription": "Ratio of number of completed demand
     load requests hitting in L2 cache to the total number of completed
     instructions",
      > +        "MetricExpr": "MEM_LOAD_UOPS_RETIRED.L2_HIT /
     INST_RETIRED.ANY",
      > +        "MetricGroup": "",
      > +        "MetricName": "l2_demand_data_read_hits_per_instr",
      > +        "ScaleUnit": "1per_instr"
      > +    },
      > +    {
      > +        "BriefDescription": "Ratio of number of requests missing
     L2 cache (includes code+data+rfo w/ prefetches) to the total number
     of completed instructions",
      > +        "MetricExpr": "L2_LINES_IN.ALL / INST_RETIRED.ANY",
      > +        "MetricGroup": "",
      > +        "MetricName":
     "l2_mpi_includes_code_plus_data_plus_rfo_with_prefetches",
      > +        "ScaleUnit": "1per_instr"
      > +    },
      > +    {
      > +        "BriefDescription": "Ratio of number of completed data
     read request missing L2 cache to the total number of completed
     instructions",
      > +        "MetricExpr": "MEM_LOAD_UOPS_RETIRED.L2_MISS /
     INST_RETIRED.ANY",
      > +        "MetricGroup": "",
      > +        "MetricName": "l2_demand_data_read_mpi",
      > +        "ScaleUnit": "1per_instr"
      > +    },
      > +    {
      > +        "BriefDescription": "Ratio of number of code read
     request missing L2 cache to the total number of completed instructions",
      > +        "MetricExpr": "L2_RQSTS.CODE_RD_MISS / INST_RETIRED.ANY",
      > +        "MetricGroup": "",
      > +        "MetricName": "l2_demand_code_mpi",
      > +        "ScaleUnit": "1per_instr"
      > +    },
      > +    {
      > +        "BriefDescription": "Ratio of number of data read
     requests missing last level core cache (includes demand w/
     prefetches) to the total number of completed instructions",
      > +        "MetricExpr": "(
     cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x182@ +
     cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x192@ ) /
     INST_RETIRED.ANY",
      > +        "MetricGroup": "",
      > +        "MetricName": "llc_data_read_mpi_demand_plus_prefetch",
      > +        "ScaleUnit": "1per_instr"
      > +    },
      > +    {
      > +        "BriefDescription": "Ratio of number of code read
     requests missing last level core cache (includes demand w/
     prefetches) to the total number of completed instructions",
      > +        "MetricExpr": "(
     cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x181@ +
     cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x191@ ) /
     INST_RETIRED.ANY",
      > +        "MetricGroup": "",
      > +        "MetricName": "llc_code_read_mpi_demand_plus_prefetch",
      > +        "ScaleUnit": "1per_instr"
      > +    },
      > +    {
      > +        "BriefDescription": "Average latency of a last level
     cache (LLC) demand and prefetch data read miss (read memory access)
     in nano seconds",
      > +        "MetricExpr": "( 1000000000 * (
     cbox@UNC_C_TOR_OCCUPANCY.MISS_OPCODE\\,filter_opc\\=0x182@ /
     cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x182@ ) / (
     UNC_C_CLOCKTICKS / ( source_count(UNC_C_CLOCKTICKS) * #num_packages
     ) ) ) * duration_time",
      > +        "MetricGroup": "",
      > +        "MetricName":
     "llc_data_read_demand_plus_prefetch_miss_latency",
      > +        "ScaleUnit": "1ns"
      > +    },
      > +    {
      > +        "BriefDescription": "Average latency of a last level
     cache (LLC) demand and prefetch data read miss (read memory access)
     addressed to local memory in nano seconds",
      > +        "MetricExpr": "( 1000000000 * (
     cbox@UNC_C_TOR_OCCUPANCY.MISS_LOCAL_OPCODE\\,filter_opc\\=0x182@ /
     cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x182@ ) / (
     UNC_C_CLOCKTICKS / ( source_count(UNC_C_CLOCKTICKS) * #num_packages
     ) ) ) * duration_time",
      > +        "MetricGroup": "",
      > +        "MetricName":
     "llc_data_read_demand_plus_prefetch_miss_latency_for_local_requests",
      > +        "ScaleUnit": "1ns"
      > +    },
      > +    {
      > +        "BriefDescription": "Average latency of a last level
     cache (LLC) demand and prefetch data read miss (read memory access)
     addressed to remote memory in nano seconds",
      > +        "MetricExpr": "( 1000000000 * (
     cbox@UNC_C_TOR_OCCUPANCY.MISS_REMOTE_OPCODE\\,filter_opc\\=0x182@ /
     cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x182@ ) / (
     UNC_C_CLOCKTICKS / ( source_count(UNC_C_CLOCKTICKS) * #num_packages
     ) ) ) * duration_time",
      > +        "MetricGroup": "",
      > +        "MetricName":
     "llc_data_read_demand_plus_prefetch_miss_latency_for_remote_requests",
      > +        "ScaleUnit": "1ns"
      > +    },
      > +    {
      > +        "BriefDescription": "Ratio of number of completed page
     walks (for all page sizes) caused by a code fetch to the total
     number of completed instructions. This implies it missed in the ITLB
     (Instruction TLB) and further levels of TLB.",
      > +        "MetricExpr": "ITLB_MISSES.WALK_COMPLETED /
     INST_RETIRED.ANY",
      > +        "MetricGroup": "",
      > +        "MetricName": "itlb_mpi",
      > +        "ScaleUnit": "1per_instr"
      > +    },
      > +    {
      > +        "BriefDescription": "Ratio of number of completed page
     walks (for 2 megabyte and 4 megabyte page sizes) caused by a code
     fetch to the total number of completed instructions. This implies it
     missed in the Instruction Translation Lookaside Buffer (ITLB) and
     further levels of TLB.",
      > +        "MetricExpr": "ITLB_MISSES.WALK_COMPLETED_2M_4M /
     INST_RETIRED.ANY",
      > +        "MetricGroup": "",
      > +        "MetricName": "itlb_large_page_mpi",
      > +        "ScaleUnit": "1per_instr"
      > +    },
      > +    {
      > +        "BriefDescription": "Ratio of number of completed page
     walks (for all page sizes) caused by demand data loads to the total
     number of completed instructions. This implies it missed in the DTLB
     and further levels of TLB.",
      > +        "MetricExpr": "DTLB_LOAD_MISSES.WALK_COMPLETED /
     INST_RETIRED.ANY",
      > +        "MetricGroup": "",
      > +        "MetricName": "dtlb_load_mpi",
      > +        "ScaleUnit": "1per_instr"
      > +    },
      > +    {
      > +        "BriefDescription": "Ratio of number of completed page
     walks (for all page sizes) caused by demand data stores to the total
     number of completed instructions. This implies it missed in the DTLB
     and further levels of TLB.",
      > +        "MetricExpr": "DTLB_STORE_MISSES.WALK_COMPLETED /
     INST_RETIRED.ANY",
      > +        "MetricGroup": "",
      > +        "MetricName": "dtlb_store_mpi",
      > +        "ScaleUnit": "1per_instr"
      > +    },
      > +    {
      > +        "BriefDescription": "Memory read that miss the last
     level cache (LLC) addressed to local DRAM as a percentage of total
     memory read accesses, does not include LLC prefetches.",
      > +        "MetricExpr": "100 *
     cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x182@ / (
     cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x182@ +
     cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x182@ )",
      > +        "MetricGroup": "",
      > +        "MetricName": "numa_percent_reads_addressed_to_local_dram",
      > +        "ScaleUnit": "1%"
      > +    },
      > +    {
      > +        "BriefDescription": "Memory reads that miss the last
     level cache (LLC) addressed to remote DRAM as a percentage of total
     memory read accesses, does not include LLC prefetches.",
      > +        "MetricExpr": "100 *
     cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x182@ / (
     cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x182@ +
     cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x182@ )",
      > +        "MetricGroup": "",
      > +        "MetricName": "numa_percent_reads_addressed_to_remote_dram",
      > +        "ScaleUnit": "1%"
      > +    },
      > +    {
      > +        "BriefDescription": "Uncore operating frequency in GHz",
      > +        "MetricExpr": "UNC_C_CLOCKTICKS / (
     source_count(UNC_C_CLOCKTICKS) * #num_packages ) / 1000000000",
      > +        "MetricGroup": "",
      > +        "MetricName": "uncore_frequency",
      > +        "ScaleUnit": "1GHz"
      > +    },
      > +    {
      > +        "BriefDescription": "Intel(R) Quick Path Interconnect
     (QPI) data transmit bandwidth (MB/sec)",
      > +        "MetricExpr": "( UNC_Q_TxL_FLITS_G0.DATA * 8 / 1000000)
     / duration_time",
      > +        "MetricGroup": "",
      > +        "MetricName": "qpi_data_transmit_bw_only_data",
      > +        "ScaleUnit": "1MB/s"
      > +    },
      > +    {
      > +        "BriefDescription": "DDR memory read bandwidth (MB/sec)",
      > +        "MetricExpr": "( UNC_M_CAS_COUNT.RD * 64 / 1000000) /
     duration_time",
      > +        "MetricGroup": "",
      > +        "MetricName": "memory_bandwidth_read",
      > +        "ScaleUnit": "1MB/s"
      > +    },
      > +    {
      > +        "BriefDescription": "DDR memory write bandwidth (MB/sec)",
      > +        "MetricExpr": "( UNC_M_CAS_COUNT.WR * 64 / 1000000) /
     duration_time",
      > +        "MetricGroup": "",
      > +        "MetricName": "memory_bandwidth_write",
      > +        "ScaleUnit": "1MB/s"
      > +    },
      > +    {
      > +        "BriefDescription": "DDR memory bandwidth (MB/sec)",
      > +        "MetricExpr": "(( UNC_M_CAS_COUNT.RD +
     UNC_M_CAS_COUNT.WR ) * 64 / 1000000) / duration_time",
      > +        "MetricGroup": "",
      > +        "MetricName": "memory_bandwidth_total",
      > +        "ScaleUnit": "1MB/s"
      > +    },
      > +    {
      > +        "BriefDescription": "Bandwidth of IO reads that are
     initiated by end device controllers that are requesting memory from
     the CPU.",
      > +        "MetricExpr": "(
     cbox@UNC_C_TOR_INSERTS.OPCODE\\,filter_opc\\=0x19e@ * 64 / 1000000)
     / duration_time",
      > +        "MetricGroup": "",
      > +        "MetricName": "io_bandwidth_read",
      > +        "ScaleUnit": "1MB/s"
      > +    },
      > +    {
      > +        "BriefDescription": "Bandwidth of IO writes that are
     initiated by end device controllers that are writing memory to the
     CPU.",
      > +        "MetricExpr": "((
     cbox@UNC_C_TOR_INSERTS.OPCODE\\,filter_opc\\=0x1c8\\,filter_tid\\=0x3e@
     +
     cbox@UNC_C_TOR_INSERTS.OPCODE\\,filter_opc\\=0x180\\,filter_tid\\=0x3e@
     ) * 64 / 1000000) / duration_time",
      > +        "MetricGroup": "",
      > +        "MetricName": "io_bandwidth_write",
      > +        "ScaleUnit": "1MB/s"
      > +    },
      > +    {
      > +        "BriefDescription": "Uops delivered from decoded
     instruction cache (decoded stream buffer or DSB) as a percent of
     total uops delivered to Instruction Decode Queue",
      > +        "MetricExpr": "100 * ( IDQ.DSB_UOPS / UOPS_ISSUED.ANY )",
      > +        "MetricGroup": "",
      > +        "MetricName":
     "percent_uops_delivered_from_decoded_icache_dsb",
      > +        "ScaleUnit": "1%"
      > +    },
      > +    {
      > +        "BriefDescription": "Uops delivered from legacy decode
     pipeline (Micro-instruction Translation Engine or MITE) as a percent
     of total uops delivered to Instruction Decode Queue",
      > +        "MetricExpr": "100 * ( IDQ.MITE_UOPS / UOPS_ISSUED.ANY )",
      > +        "MetricGroup": "",
      > +        "MetricName":
     "percent_uops_delivered_from_legacy_decode_pipeline_mite",
      > +        "ScaleUnit": "1%"
      > +    },
      > +    {
      > +        "BriefDescription": "Uops delivered from microcode
     sequencer (MS) as a percent of total uops delivered to Instruction
     Decode Queue",
      > +        "MetricExpr": "100 * ( IDQ.MS_UOPS / UOPS_ISSUED.ANY )",
      > +        "MetricGroup": "",
      > +        "MetricName":
     "percent_uops_delivered_from_microcode_sequencer_ms",
      > +        "ScaleUnit": "1%"
      > +    },
      > +    {
      > +        "BriefDescription": "Uops delivered from loop stream
     detector(LSD) as a percent of total uops delivered to Instruction
     Decode Queue",
      > +        "MetricExpr": "100 * ( LSD.UOPS / UOPS_ISSUED.ANY )",
      > +        "MetricGroup": "",
      > +        "MetricName":
     "percent_uops_delivered_from_loop_stream_detector_lsd",
      > +        "ScaleUnit": "1%"
      > +    },
      > +    {
      > +        "BriefDescription": "This category represents fraction
     of slots where the processor's Frontend undersupplies its Backend.
     Frontend denotes the first part of the processor core responsible to
     fetch operations that are executed later on by the Backend part.
     Within the Frontend; a branch predictor predicts the next address to
     fetch; cache-lines are fetched from the memory subsystem; parsed
     into instructions; and lastly decoded into micro-operations (uops).
     Ideally the Frontend can issue Machine_Width uops every cycle to the
     Backend. Frontend Bound denotes unutilized issue-slots when there is
     no Backend stall; i.e. bubbles where Frontend delivered no uops
     while Backend could have accepted them. For example; stalls due to
     instruction-cache misses would be categorized under Frontend Bound.",
      > +        "MetricExpr": "100 * ( IDQ_UOPS_NOT_DELIVERED.CORE / ( (
     4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else (
     CPU_CLK_UNHALTED.THREAD ) ) ) )",
      > +        "MetricGroup": "TmaL1;PGO",
      > +        "MetricName": "tma_frontend_bound_percent",
      > +        "ScaleUnit": "1%"
      > +    },
      > +    {
      > +        "BriefDescription": "This metric represents fraction of
     slots the CPU was stalled due to Frontend latency issues.  For
     example; instruction-cache misses; iTLB misses or fetch stalls after
     a branch misprediction are categorized under Frontend Latency. In
     such cases; the Frontend eventually delivers no uops for some period.",
      > +        "MetricExpr": "100 * ( ( 4 ) *
     IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / ( ( 4 ) * ( (
     CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else (
     CPU_CLK_UNHALTED.THREAD ) ) ) )",
      > +        "MetricGroup":
     "Frontend;TmaL2;m_tma_frontend_bound_percent",
      > +        "MetricName": "tma_fetch_latency_percent",
      > +        "ScaleUnit": "1%"
      > +    },
      > +    {
      > +        "BriefDescription": "This metric represents fraction of
     cycles the CPU was stalled due to instruction cache misses.",
      > +        "MetricExpr": "100 * ( ICACHE.IFDATA_STALL / (
     CPU_CLK_UNHALTED.THREAD ) )",
      > +        "MetricGroup":
     "BigFoot;FetchLat;IcMiss;TmaL3;m_tma_fetch_latency_percent",
      > +        "MetricName": "tma_icache_misses_percent",
      > +        "ScaleUnit": "1%"
      > +    },
      > +    {
      > +        "BriefDescription": "This metric represents fraction of
     cycles the CPU was stalled due to Instruction TLB (ITLB) misses.",
      > +        "MetricExpr": "100 * ( ( 14 * ITLB_MISSES.STLB_HIT +
     cpu@ITLB_MISSES.WALK_DURATION\\,cmask\\=0x1@ + 7 *
     ITLB_MISSES.WALK_COMPLETED ) / ( CPU_CLK_UNHALTED.THREAD ) )",
      > +        "MetricGroup":
     "BigFoot;FetchLat;MemoryTLB;TmaL3;m_tma_fetch_latency_percent",
      > +        "MetricName": "tma_itlb_misses_percent",
      > +        "ScaleUnit": "1%"
      > +    },
      > +    {
      > +        "BriefDescription": "This metric represents fraction of
     cycles the CPU was stalled due to Branch Resteers. Branch Resteers
     estimates the Frontend delay in fetching operations from corrected
     path; following all sorts of miss-predicted branches. For example;
     branchy code with lots of miss-predictions might get categorized
     under Branch Resteers. Note the value of this node may overlap with
     its siblings.",
      > +        "MetricExpr": "100 * ( ( 12 ) * (
     BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT + BACLEARS.ANY )
     / ( CPU_CLK_UNHALTED.THREAD ) )",
      > +        "MetricGroup": "FetchLat;TmaL3;m_tma_fetch_latency_percent",
      > +        "MetricName": "tma_branch_resteers_percent",
      > +        "ScaleUnit": "1%"
      > +    },
      > +    {
      > +        "BriefDescription": "This metric represents fraction of
     cycles the CPU was stalled due to switches from DSB to MITE
     pipelines. The DSB (decoded i-cache) is a Uop Cache where the
     front-end directly delivers Uops (micro operations) avoiding heavy
     x86 decoding. The DSB pipeline has shorter latency and delivered
     higher bandwidth than the MITE (legacy instruction decode pipeline).
     Switching between the two pipelines can cause penalties hence this
     metric measures the exposed penalty.",
      > +        "MetricExpr": "100 * ( DSB2MITE_SWITCHES.PENALTY_CYCLES
     / ( CPU_CLK_UNHALTED.THREAD ) )",
      > +        "MetricGroup":
     "DSBmiss;FetchLat;TmaL3;m_tma_fetch_latency_percent",
      > +        "MetricName": "tma_dsb_switches_percent",
      > +        "ScaleUnit": "1%"
      > +    },
      > +    {
      > +        "BriefDescription": "This metric represents fraction of
     cycles CPU was stalled due to Length Changing Prefixes (LCPs). Using
     proper compiler flags or Intel Compiler by default will certainly
     avoid this. #Link: Optimization Guide about LCP BKMs.",
      > +        "MetricExpr": "100 * ( ILD_STALL.LCP / (
     CPU_CLK_UNHALTED.THREAD ) )",
      > +        "MetricGroup": "FetchLat;TmaL3;m_tma_fetch_latency_percent",
      > +        "MetricName": "tma_lcp_percent",
      > +        "ScaleUnit": "1%"
      > +    },
      > +    {
      > +        "BriefDescription": "This metric estimates the fraction
     of cycles when the CPU was stalled due to switches of uop delivery
     to the Microcode Sequencer (MS). Commonly used instructions are
     optimized for delivery by the DSB (decoded i-cache) or MITE (legacy
     instruction decode) pipelines. Certain operations cannot be handled
     natively by the execution pipeline; and must be performed by
     microcode (small programs injected into the execution stream).
     Switching to the MS too often can negatively impact performance. The
     MS is designated to deliver long uop flows required by CISC
     instructions like CPUID; or uncommon conditions like Floating Point
     Assists when dealing with Denormals.",
      > +        "MetricExpr": "100 * ( ( 2 ) * IDQ.MS_SWITCHES / (
     CPU_CLK_UNHALTED.THREAD ) )",
      > +        "MetricGroup":
     "FetchLat;MicroSeq;TmaL3;m_tma_fetch_latency_percent",
      > +        "MetricName": "tma_ms_switches_percent",
      > +        "ScaleUnit": "1%"
      > +    },
      > +    {
      > +        "BriefDescription": "This metric represents fraction of
     slots the CPU was stalled due to Frontend bandwidth issues.  For
     example; inefficiencies at the instruction decoders; or restrictions
     for caching in the DSB (decoded uops cache) are categorized under
     Fetch Bandwidth. In such cases; the Frontend typically delivers
     suboptimal amount of uops to the Backend.",
      > +        "MetricExpr": "100 * ( ( IDQ_UOPS_NOT_DELIVERED.CORE / (
     ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else (
     CPU_CLK_UNHALTED.THREAD ) ) ) ) - ( ( 4 ) *
     IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / ( ( 4 ) * ( (
     CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else (
     CPU_CLK_UNHALTED.THREAD ) ) ) ) )",
      > +        "MetricGroup":
     "FetchBW;Frontend;TmaL2;m_tma_frontend_bound_percent",
      > +        "MetricName": "tma_fetch_bandwidth_percent",
      > +        "ScaleUnit": "1%"
      > +    },
      > +    {
      > +        "BriefDescription": "This metric represents Core
     fraction of cycles in which CPU was likely limited due to the MITE
     pipeline (the legacy decode pipeline). This pipeline is used for
     code that was not pre-cached in the DSB or LSD. For example;
     inefficiencies due to asymmetric decoders; use of long immediate or
     LCP can manifest as MITE fetch bandwidth bottleneck.",
      > +        "MetricExpr": "100 * ( ( IDQ.ALL_MITE_CYCLES_ANY_UOPS -
     IDQ.ALL_MITE_CYCLES_4_UOPS ) / ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 )
     if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) / 2 )",
      > +        "MetricGroup":
     "DSBmiss;FetchBW;TmaL3;m_tma_fetch_bandwidth_percent",
      > +        "MetricName": "tma_mite_percent",
      > +        "ScaleUnit": "1%"
      > +    },
      > +    {
      > +        "BriefDescription": "This metric represents Core
     fraction of cycles in which CPU was likely limited due to DSB
     (decoded uop cache) fetch pipeline.  For example; inefficient
     utilization of the DSB cache structure or bank conflict when reading
     from it; are categorized here.",
      > +        "MetricExpr": "100 * ( ( IDQ.ALL_DSB_CYCLES_ANY_UOPS -
     IDQ.ALL_DSB_CYCLES_4_UOPS ) / ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 )
     if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) / 2 )",
      > +        "MetricGroup":
     "DSB;FetchBW;TmaL3;m_tma_fetch_bandwidth_percent",
      > +        "MetricName": "tma_dsb_percent",
      > +        "ScaleUnit": "1%"
      > +    },
      > +    {
      > +        "BriefDescription": "This category represents fraction
     of slots wasted due to incorrect speculations. This include slots
     used to issue uops that do not eventually get retired and slots for
     which the issue-pipeline was blocked due to recovery from earlier
     incorrect speculation. For example; wasted work due to
     miss-predicted branches are categorized under Bad Speculation
     category. Incorrect data speculation followed by Memory Ordering
     Nukes is another example.",
      > +        "MetricExpr": "100 * ( ( UOPS_ISSUED.ANY - (
     UOPS_RETIRED.RETIRE_SLOTS ) + ( 4 ) * ( (
     INT_MISC.RECOVERY_CYCLES_ANY / 2 ) if #SMT_on else
     INT_MISC.RECOVERY_CYCLES ) ) / ( ( 4 ) * ( (
     CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else (
     CPU_CLK_UNHALTED.THREAD ) ) ) )",
      > +        "MetricGroup": "TmaL1",
      > +        "MetricName": "tma_bad_speculation_percent",
      > +        "ScaleUnit": "1%"
      > +    },
      > +    {
      > +        "BriefDescription": "This metric represents fraction of
     slots the CPU has wasted due to Branch Misprediction.  These slots
     are either wasted by uops fetched from an incorrectly speculated
     program path; or stalls when the out-of-order part of the machine
     needs to recover its state from a speculative path.",
      > +        "MetricExpr": "100 * ( ( BR_MISP_RETIRED.ALL_BRANCHES /
     ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT ) ) * ( (
     UOPS_ISSUED.ANY - ( UOPS_RETIRED.RETIRE_SLOTS ) + ( 4 ) * ( (
     INT_MISC.RECOVERY_CYCLES_ANY / 2 ) if #SMT_on else
     INT_MISC.RECOVERY_CYCLES ) ) / ( ( 4 ) * ( (
     CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else (
     CPU_CLK_UNHALTED.THREAD ) ) ) ) )",
      > +        "MetricGroup":
     "BadSpec;BrMispredicts;TmaL2;m_tma_bad_speculation_percent",
      > +        "MetricName": "tma_branch_mispredicts_percent",
      > +        "ScaleUnit": "1%"
      > +    },
      > +    {
      > +        "BriefDescription": "This metric represents fraction of
     slots the CPU has wasted due to Machine Clears.  These slots are
     either wasted by uops fetched prior to the clear; or stalls the
     out-of-order portion of the machine needs to recover its state after
     the clear. For example; this can happen due to memory ordering Nukes
     (e.g. Memory Disambiguation) or Self-Modifying-Code (SMC) nukes.",
      > +        "MetricExpr": "100 * ( ( ( UOPS_ISSUED.ANY - (
     UOPS_RETIRED.RETIRE_SLOTS ) + ( 4 ) * ( (
     INT_MISC.RECOVERY_CYCLES_ANY / 2 ) if #SMT_on else
     INT_MISC.RECOVERY_CYCLES ) ) / ( ( 4 ) * ( (
     CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else (
     CPU_CLK_UNHALTED.THREAD ) ) ) ) - ( ( BR_MISP_RETIRED.ALL_BRANCHES /
     ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT ) ) * ( (
     UOPS_ISSUED.ANY - ( UOPS_RETIRED.RETIRE_SLOTS ) + ( 4 ) * ( (
     INT_MISC.RECOVERY_CYCLES_ANY / 2 ) if #SMT_on else
     INT_MISC.RECOVERY_CYCLES ) ) / ( ( 4 ) * ( (
     CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else (
     CPU_CLK_UNHALTED.THREAD ) ) ) ) ) )",
      > +        "MetricGroup":
     "BadSpec;MachineClears;TmaL2;m_tma_bad_speculation_percent",
      > +        "MetricName": "tma_machine_clears_percent",
      > +        "ScaleUnit": "1%"
      > +    },
      > +    {
      > +        "BriefDescription": "This category represents fraction
     of slots where no uops are being delivered due to a lack of required
     resources for accepting new uops in the Backend. Backend is the
     portion of the processor core where the out-of-order scheduler
     dispatches ready uops into their respective execution units; and
     once completed these uops get retired according to program order.
     For example; stalls due to data-cache misses or stalls due to the
     divider unit being overloaded are both categorized under Backend
     Bound. Backend Bound is further divided into two main categories:
     Memory Bound and Core Bound.",
      > +        "MetricExpr": "100 * ( 1 - ( (
     IDQ_UOPS_NOT_DELIVERED.CORE / ( ( 4 ) * ( (
     CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else (
     CPU_CLK_UNHALTED.THREAD ) ) ) ) + ( ( UOPS_ISSUED.ANY - (
     UOPS_RETIRED.RETIRE_SLOTS ) + ( 4 ) * ( (
     INT_MISC.RECOVERY_CYCLES_ANY / 2 ) if #SMT_on else
     INT_MISC.RECOVERY_CYCLES ) ) / ( ( 4 ) * ( (
     CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else (
     CPU_CLK_UNHALTED.THREAD ) ) ) ) + ( ( UOPS_RETIRED.RETIRE_SLOTS ) /
     ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else (
     CPU_CLK_UNHALTED.THREAD ) ) ) ) ) )",
      > +        "MetricGroup": "TmaL1",
      > +        "MetricName": "tma_backend_bound_percent",
      > +        "ScaleUnit": "1%"
      > +    },
      > +    {
      > +        "BriefDescription": "This metric represents fraction of
     slots the Memory subsystem within the Backend was a bottleneck.
     Memory Bound estimates fraction of slots where pipeline is likely
     stalled due to demand load or store instructions. This accounts
     mainly for (1) non-completed in-flight memory demand loads which
     coincides with execution units starvation; in addition to (2) cases
     where stores could impose backpressure on the pipeline when many of
     them get buffered at the same time (less common out of the two).",
      > +        "MetricExpr": "100 * ( ( ( CYCLE_ACTIVITY.STALLS_MEM_ANY
     + RESOURCE_STALLS.SB <http://RESOURCE_STALLS.SB> ) / ( (
     CYCLE_ACTIVITY.STALLS_TOTAL + UOPS_EXECUTED.CYCLES_GE_1_UOP_EXEC - (
     UOPS_EXECUTED.CYCLES_GE_3_UOPS_EXEC if ( ( INST_RETIRED.ANY / (
     CPU_CLK_UNHALTED.THREAD ) ) > 1.8 ) else
     UOPS_EXECUTED.CYCLES_GE_2_UOPS_EXEC ) - ( RS_EVENTS.EMPTY_CYCLES if
     ( ( ( 4 ) * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / ( ( 4
     ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else (
     CPU_CLK_UNHALTED.THREAD ) ) ) ) > 0.1 ) else 0 ) +
     RESOURCE_STALLS.SB <http://RESOURCE_STALLS.SB> ) ) ) * ( 1 - ( (
     IDQ_UOPS_NOT_DELIVERED.CORE / ( ( 4 ) * ( (
     CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else (
     CPU_CLK_UNHALTED.THREAD ) ) ) ) + ( ( UOPS_ISSUED.ANY - (
     UOPS_RETIRED.RETIRE_SLOTS ) + ( 4 ) * ( (
     INT_MISC.RECOVERY_CYCLES_ANY / 2 ) if #SMT_on else
     INT_MISC.RECOVERY_CYCLES ) ) / ( ( 4 ) * ( (
     CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else (
     CPU_CLK_UNHALTED.THREAD ) ) ) ) + ( ( UOPS_RETIRED.RETIRE_SLOTS ) /
     ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else (
     CPU_CLK_UNHALTED.THREAD ) ) ) ) ) ) )",
      > +        "MetricGroup": "Backend;TmaL2;m_tma_backend_bound_percent",
      > +        "MetricName": "tma_memory_bound_percent",
      > +        "ScaleUnit": "1%"
      > +    },
      > +    {
      > +        "BriefDescription": "This metric estimates how often the
     CPU was stalled without loads missing the L1 data cache.  The L1
     data cache typically has the shortest latency.  However; in certain
     cases like loads blocked on older stores; a load might suffer due to
     high latency even though it is being satisfied by the L1. Another
     example is loads who miss in the TLB. These cases are characterized
     by execution unit stalls; while some non-completed demand load lives
     in the machine without having that demand load missing the L1 cache.",
      > +        "MetricExpr": "100 * ( max( (
     CYCLE_ACTIVITY.STALLS_MEM_ANY - CYCLE_ACTIVITY.STALLS_L1D_MISS ) / (
     CPU_CLK_UNHALTED.THREAD ) , 0 ) )",
      > +        "MetricGroup":
     "CacheMisses;MemoryBound;TmaL3mem;TmaL3;m_tma_memory_bound_percent",
      > +        "MetricName": "tma_l1_bound_percent",
      > +        "ScaleUnit": "1%"
      > +    },
      > +    {
      > +        "BriefDescription": "This metric estimates how often the
     CPU was stalled due to L2 cache accesses by loads.  Avoiding cache
     misses (i.e. L1 misses/L2 hits) can improve the latency and increase
     performance.",
      > +        "MetricExpr": "100 * ( ( CYCLE_ACTIVITY.STALLS_L1D_MISS
     - CYCLE_ACTIVITY.STALLS_L2_MISS ) / ( CPU_CLK_UNHALTED.THREAD ) )",
      > +        "MetricGroup":
     "CacheMisses;MemoryBound;TmaL3mem;TmaL3;m_tma_memory_bound_percent",
      > +        "MetricName": "tma_l2_bound_percent",
      > +        "ScaleUnit": "1%"
      > +    },
      > +    {
      > +        "BriefDescription": "This metric estimates how often the
     CPU was stalled due to loads accesses to L3 cache or contended with
     a sibling Core.  Avoiding cache misses (i.e. L2 misses/L3 hits) can
     improve the latency and increase performance.",
      > +        "MetricExpr": "100 * ( ( MEM_LOAD_UOPS_RETIRED.L3_HIT /
     ( MEM_LOAD_UOPS_RETIRED.L3_HIT + ( 7 ) *
     MEM_LOAD_UOPS_RETIRED.L3_MISS ) ) * CYCLE_ACTIVITY.STALLS_L2_MISS /
     ( CPU_CLK_UNHALTED.THREAD ) )",
      > +        "MetricGroup":
     "CacheMisses;MemoryBound;TmaL3mem;TmaL3;m_tma_memory_bound_percent",
      > +        "MetricName": "tma_l3_bound_percent",
      > +        "ScaleUnit": "1%"
      > +    },
      > +    {
      > +        "BriefDescription": "This metric estimates how often the
     CPU was stalled on accesses to external memory (DRAM) by loads.
     Better caching can improve the latency and increase performance.",
      > +        "MetricExpr": "100 * ( min( ( ( 1 - (
     MEM_LOAD_UOPS_RETIRED.L3_HIT / ( MEM_LOAD_UOPS_RETIRED.L3_HIT + ( 7
     ) * MEM_LOAD_UOPS_RETIRED.L3_MISS ) ) ) *
     CYCLE_ACTIVITY.STALLS_L2_MISS / ( CPU_CLK_UNHALTED.THREAD ) ) , ( 1
     ) ) )",
      > +        "MetricGroup":
     "MemoryBound;TmaL3mem;TmaL3;m_tma_memory_bound_percent",
      > +        "MetricName": "tma_dram_bound_percent",
      > +        "ScaleUnit": "1%"
      > +    },
      > +    {
      > +        "BriefDescription": "This metric estimates how often CPU
     was stalled  due to RFO store memory accesses; RFO store issue a
     read-for-ownership request before the write. Even though store
     accesses do not typically stall out-of-order CPUs; there are few
     cases where stores can lead to actual stalls. This metric will be
     flagged should RFO stores be a bottleneck.",
      > +        "MetricExpr": "100 * ( RESOURCE_STALLS.SB
     <http://RESOURCE_STALLS.SB> / ( CPU_CLK_UNHALTED.THREAD ) )",
      > +        "MetricGroup":
     "MemoryBound;TmaL3mem;TmaL3;m_tma_memory_bound_percent",
      > +        "MetricName": "tma_store_bound_percent",
      > +        "ScaleUnit": "1%"
      > +    },
      > +    {
      > +        "BriefDescription": "This metric represents fraction of
     slots where Core non-memory issues were of a bottleneck.  Shortage
     in hardware compute resources; or dependencies in software's
     instructions are both categorized under Core Bound. Hence it may
     indicate the machine ran out of an out-of-order resource; certain
     execution units are overloaded or dependencies in program's data- or
     instruction-flow are limiting the performance (e.g. FP-chained
     long-latency arithmetic operations).",
      > +        "MetricExpr": "100 * ( ( 1 - ( (
     IDQ_UOPS_NOT_DELIVERED.CORE / ( ( 4 ) * ( (
     CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else (
     CPU_CLK_UNHALTED.THREAD ) ) ) ) + ( ( UOPS_ISSUED.ANY - (
     UOPS_RETIRED.RETIRE_SLOTS ) + ( 4 ) * ( (
     INT_MISC.RECOVERY_CYCLES_ANY / 2 ) if #SMT_on else
     INT_MISC.RECOVERY_CYCLES ) ) / ( ( 4 ) * ( (
     CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else (
     CPU_CLK_UNHALTED.THREAD ) ) ) ) + ( ( UOPS_RETIRED.RETIRE_SLOTS ) /
     ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else (
     CPU_CLK_UNHALTED.THREAD ) ) ) ) ) ) - ( ( (
     CYCLE_ACTIVITY.STALLS_MEM_ANY + RESOURCE_STALLS.SB
     <http://RESOURCE_STALLS.SB> ) / ( ( CYCLE_ACTIVITY.STALLS_TOTAL +
     UOPS_EXECUTED.CYCLES_GE_1_UOP_EXEC - (
     UOPS_EXECUTED.CYCLES_GE_3_UOPS_EXEC if ( ( INST_RETIRED.ANY / (
     CPU_CLK_UNHALTED.THREAD ) ) > 1.8 ) else
     UOPS_EXECUTED.CYCLES_GE_2_UOPS_EXEC ) - ( RS_EVENTS.EMPTY_CYCLES if
     ( ( ( 4 ) * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / ( ( 4
     ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else (
     CPU_CLK_UNHALTED.THREAD ) ) ) ) > 0.1 ) else 0 ) +
     RESOURCE_STALLS.SB <http://RESOURCE_STALLS.SB> ) ) ) * ( 1 - ( (
     IDQ_UOPS_NOT_DELIVERED.CORE / ( ( 4 ) * ( (
     CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else (
     CPU_CLK_UNHALTED.THREAD ) ) ) ) + ( ( UOPS_ISSUED.ANY - (
     UOPS_RETIRED.RETIRE_SLOTS ) + ( 4 ) * ( (
     INT_MISC.RECOVERY_CYCLES_ANY / 2 ) if #SMT_on else
     INT_MISC.RECOVERY_CYCLES ) ) / ( ( 4 ) * ( (
     CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else (
     CPU_CLK_UNHALTED.THREAD ) ) ) ) + ( ( UOPS_RETIRED.RETIRE_SLOTS ) /
     ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else (
     CPU_CLK_UNHALTED.THREAD ) ) ) ) ) ) ) )",
      > +        "MetricGroup":
     "Backend;TmaL2;Compute;m_tma_backend_bound_percent",
      > +        "MetricName": "tma_core_bound_percent",
      > +        "ScaleUnit": "1%"
      > +    },
      > +    {
      > +        "BriefDescription": "This metric represents fraction of
     cycles where the Divider unit was active. Divide and square root
     instructions are performed by the Divider unit and can take
     considerably longer latency than integer or Floating Point addition;
     subtraction; or multiplication.",
      > +        "MetricExpr": "100 * ( ARITH.FPU_DIV_ACTIVE / ( (
     CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else (
     CPU_CLK_UNHALTED.THREAD ) ) )",
      > +        "MetricGroup": "TmaL3;m_tma_core_bound_percent",
      > +        "MetricName": "tma_divider_percent",
      > +        "ScaleUnit": "1%"
      > +    },
      > +    {
      > +        "BriefDescription": "This metric estimates fraction of
     cycles the CPU performance was potentially limited due to Core
     computation issues (non divider-related).  Two distinct categories
     can be attributed into this metric: (1) heavy data-dependency among
     contiguous instructions would manifest in this metric - such cases
     are often referred to as low Instruction Level Parallelism (ILP).
     (2) Contention on some hardware execution unit other than Divider.
     For example; when there are too many multiply operations.",
      > +        "MetricExpr": "100 * ( ( ( ( CYCLE_ACTIVITY.STALLS_TOTAL
     + UOPS_EXECUTED.CYCLES_GE_1_UOP_EXEC - (
     UOPS_EXECUTED.CYCLES_GE_3_UOPS_EXEC if ( ( INST_RETIRED.ANY / (
     CPU_CLK_UNHALTED.THREAD ) ) > 1.8 ) else
     UOPS_EXECUTED.CYCLES_GE_2_UOPS_EXEC ) - ( RS_EVENTS.EMPTY_CYCLES if
     ( ( ( 4 ) * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / ( ( 4
     ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else (
     CPU_CLK_UNHALTED.THREAD ) ) ) ) > 0.1 ) else 0 ) +
     RESOURCE_STALLS.SB <http://RESOURCE_STALLS.SB> ) ) -
     RESOURCE_STALLS.SB <http://RESOURCE_STALLS.SB> -
     CYCLE_ACTIVITY.STALLS_MEM_ANY ) / ( CPU_CLK_UNHALTED.THREAD ) )",
      > +        "MetricGroup": "PortsUtil;TmaL3;m_tma_core_bound_percent",
      > +        "MetricName": "tma_ports_utilization_percent",
      > +        "ScaleUnit": "1%"
      > +    },
      > +    {
      > +        "BriefDescription": "This category represents fraction
     of slots utilized by useful work i.e. issued uops that eventually
     get retired. Ideally; all pipeline slots would be attributed to the
     Retiring category.  Retiring of 100% would indicate the maximum
     Pipeline_Width throughput was achieved.  Maximizing Retiring
     typically increases the Instructions-per-cycle (see IPC metric).
     Note that a high Retiring value does not necessary mean there is no
     room for more performance.  For example; Heavy-operations or
     Microcode Assists are categorized under Retiring. They often
     indicate suboptimal performance and can often be optimized or
     avoided. ",
      > +        "MetricExpr": "100 * ( ( UOPS_RETIRED.RETIRE_SLOTS ) / (
     ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else (
     CPU_CLK_UNHALTED.THREAD ) ) ) )",
      > +        "MetricGroup": "TmaL1",
      > +        "MetricName": "tma_retiring_percent",
      > +        "ScaleUnit": "1%"
      > +    },
      > +    {
      > +        "BriefDescription": "This metric represents fraction of
     slots where the CPU was retiring light-weight operations --
     instructions that require no more than one uop (micro-operation).
     This correlates with total number of instructions used by the
     program. A uops-per-instruction (see UPI metric) ratio of 1 or less
     should be expected for decently optimized software running on Intel
     Core/Xeon products. While this often indicates efficient X86
     instructions were executed; high value does not necessarily mean
     better performance cannot be achieved.",
      > +        "MetricExpr": "100 * ( ( ( UOPS_RETIRED.RETIRE_SLOTS ) /
     ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else (
     CPU_CLK_UNHALTED.THREAD ) ) ) ) - ( ( ( ( UOPS_RETIRED.RETIRE_SLOTS
     ) / UOPS_ISSUED.ANY ) * IDQ.MS_UOPS / ( ( 4 ) * ( (
     CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else (
     CPU_CLK_UNHALTED.THREAD ) ) ) ) ) )",
      > +        "MetricGroup": "Retire;TmaL2;m_tma_retiring_percent",
      > +        "MetricName": "tma_light_operations_percent",
      > +        "ScaleUnit": "1%"
      > +    },
      > +    {
      > +        "BriefDescription": "This metric represents overall
     arithmetic floating-point (FP) operations fraction the CPU has
     executed (retired). Note this metric's value may exceed its parent
     due to use of \"Uops\" CountDomain and FMA double-counting.",
      > +        "MetricExpr": "100 * ( ( INST_RETIRED.X87 * ( (
     UOPS_RETIRED.RETIRE_SLOTS ) / INST_RETIRED.ANY ) / (
     UOPS_RETIRED.RETIRE_SLOTS ) ) + ( (
     FP_ARITH_INST_RETIRED.SCALAR_SINGLE +
     FP_ARITH_INST_RETIRED.SCALAR_DOUBLE ) / ( UOPS_RETIRED.RETIRE_SLOTS
     ) ) + ( min( ( ( FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE +
     FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE +
     FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE +
     FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE ) / (
     UOPS_RETIRED.RETIRE_SLOTS ) ) , ( 1 ) ) ) )",
      > +        "MetricGroup": "HPC;TmaL3;m_tma_light_operations_percent",
      > +        "MetricName": "tma_fp_arith_percent",
      > +        "ScaleUnit": "1%"
      > +    },
      > +    {
      > +        "BriefDescription": "This metric represents fraction of
     slots where the CPU was retiring heavy-weight operations --
     instructions that require two or more uops or microcoded sequences.
     This highly-correlates with the uop length of these
     instructions/sequences.",
      > +        "MetricExpr": "100 * ( ( ( ( UOPS_RETIRED.RETIRE_SLOTS )
     / UOPS_ISSUED.ANY ) * IDQ.MS_UOPS / ( ( 4 ) * ( (
     CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else (
     CPU_CLK_UNHALTED.THREAD ) ) ) ) )",
      > +        "MetricGroup": "Retire;TmaL2;m_tma_retiring_percent",
      > +        "MetricName": "tma_heavy_operations_percent",
      > +        "ScaleUnit": "1%"
      > +    },
      > +    {
      > +        "BriefDescription": "This metric represents fraction of
     slots the CPU was retiring uops fetched by the Microcode Sequencer
     (MS) unit.  The MS is used for CISC instructions not supported by
     the default decoders (like repeat move strings; or CPUID); or by
     microcode assists used to address some operation modes (like in
     Floating Point assists). These cases can often be avoided.",
      > +        "MetricExpr": "100 * ( ( ( UOPS_RETIRED.RETIRE_SLOTS ) /
     UOPS_ISSUED.ANY ) * IDQ.MS_UOPS / ( ( 4 ) * ( (
     CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else (
     CPU_CLK_UNHALTED.THREAD ) ) ) )",
      > +        "MetricGroup":
     "MicroSeq;TmaL3;m_tma_heavy_operations_percent",
      > +        "MetricName": "tma_microcode_sequencer_percent",
      > +        "ScaleUnit": "1%"
      >       }
      >   ]
      > diff --git a/tools/perf/pmu-events/arch/x86/broadwellx/cache.json
     b/tools/perf/pmu-events/arch/x86/broadwellx/cache.json
      > index 127abe08362f..2efc4c0ee740 100644
      > --- a/tools/perf/pmu-events/arch/x86/broadwellx/cache.json
      > +++ b/tools/perf/pmu-events/arch/x86/broadwellx/cache.json
      > @@ -814,9 +814,8 @@
      >           "EventCode": "0xB7, 0xBB",
      >           "EventName":
     "OFFCORE_RESPONSE.ALL_CODE_RD.LLC_HIT.HIT_OTHER_CORE_NO_FWD",
      >           "MSRIndex": "0x1a6,0x1a7",
      > -        "MSRValue": "0x04003C0244",
      > +        "MSRValue": "0x4003C0244",
      >           "Offcore": "1",
      > -        "PublicDescription": "Counts all demand & prefetch code
     reads hit in the L3 and the snoops to sibling cores hit in either
     E/S state and the line is not forwarded",
      >           "SampleAfterValue": "100003",
      >           "UMask": "0x1"
      >       },
      > @@ -829,7 +828,6 @@
      >           "MSRIndex": "0x1a6,0x1a7",
      >           "MSRValue": "0x10003C0091",
      >           "Offcore": "1",
      > -        "PublicDescription": "Counts all demand & prefetch data
     reads hit in the L3 and the snoop to one of the sibling cores hits
     the line in M state and the line is forwarded",
      >           "SampleAfterValue": "100003",
      >           "UMask": "0x1"
      >       },
      > @@ -840,9 +838,8 @@
      >           "EventCode": "0xB7, 0xBB",
      >           "EventName":
     "OFFCORE_RESPONSE.ALL_DATA_RD.LLC_HIT.HIT_OTHER_CORE_NO_FWD",
      >           "MSRIndex": "0x1a6,0x1a7",
      > -        "MSRValue": "0x04003C0091",
      > +        "MSRValue": "0x4003C0091",
      >           "Offcore": "1",
      > -        "PublicDescription": "Counts all demand & prefetch data
     reads hit in the L3 and the snoops to sibling cores hit in either
     E/S state and the line is not forwarded",
      >           "SampleAfterValue": "100003",
      >           "UMask": "0x1"
      >       },
      > @@ -855,7 +852,6 @@
      >           "MSRIndex": "0x1a6,0x1a7",
      >           "MSRValue": "0x10003C07F7",
      >           "Offcore": "1",
      > -        "PublicDescription": "Counts all data/code/rfo reads
     (demand & prefetch) hit in the L3 and the snoop to one of the
     sibling cores hits the line in M state and the line is forwarded",
      >           "SampleAfterValue": "100003",
      >           "UMask": "0x1"
      >       },
      > @@ -866,9 +862,8 @@
      >           "EventCode": "0xB7, 0xBB",
      >           "EventName":
     "OFFCORE_RESPONSE.ALL_READS.LLC_HIT.HIT_OTHER_CORE_NO_FWD",
      >           "MSRIndex": "0x1a6,0x1a7",
      > -        "MSRValue": "0x04003C07F7",
      > +        "MSRValue": "0x4003C07F7",
      >           "Offcore": "1",
      > -        "PublicDescription": "Counts all data/code/rfo reads
     (demand & prefetch) hit in the L3 and the snoops to sibling cores
     hit in either E/S state and the line is not forwarded",
      >           "SampleAfterValue": "100003",
      >           "UMask": "0x1"
      >       },
      > @@ -881,7 +876,6 @@
      >           "MSRIndex": "0x1a6,0x1a7",
      >           "MSRValue": "0x3F803C8FFF",
      >           "Offcore": "1",
      > -        "PublicDescription": "Counts all requests hit in the L3",
      >           "SampleAfterValue": "100003",
      >           "UMask": "0x1"
      >       },
      > @@ -894,7 +888,6 @@
      >           "MSRIndex": "0x1a6,0x1a7",
      >           "MSRValue": "0x10003C0122",
      >           "Offcore": "1",
      > -        "PublicDescription": "Counts all demand & prefetch RFOs
     hit in the L3 and the snoop to one of the sibling cores hits the
     line in M state and the line is forwarded",
      >           "SampleAfterValue": "100003",
      >           "UMask": "0x1"
      >       },
      > @@ -905,9 +898,8 @@
      >           "EventCode": "0xB7, 0xBB",
      >           "EventName":
     "OFFCORE_RESPONSE.ALL_RFO.LLC_HIT.HIT_OTHER_CORE_NO_FWD",
      >           "MSRIndex": "0x1a6,0x1a7",
      > -        "MSRValue": "0x04003C0122",
      > +        "MSRValue": "0x4003C0122",
      >           "Offcore": "1",
      > -        "PublicDescription": "Counts all demand & prefetch RFOs
     hit in the L3 and the snoops to sibling cores hit in either E/S
     state and the line is not forwarded",
      >           "SampleAfterValue": "100003",
      >           "UMask": "0x1"
      >       },
      > @@ -920,7 +912,6 @@
      >           "MSRIndex": "0x1a6,0x1a7",
      >           "MSRValue": "0x3F803C0002",
      >           "Offcore": "1",
      > -        "PublicDescription": "Counts all demand data writes
     (RFOs) hit in the L3",
      >           "SampleAfterValue": "100003",
      >           "UMask": "0x1"
      >       },
      > @@ -933,7 +924,6 @@
      >           "MSRIndex": "0x1a6,0x1a7",
      >           "MSRValue": "0x10003C0002",
      >           "Offcore": "1",
      > -        "PublicDescription": "Counts all demand data writes
     (RFOs) hit in the L3 and the snoop to one of the sibling cores hits
     the line in M state and the line is forwarded",
      >           "SampleAfterValue": "100003",
      >           "UMask": "0x1"
      >       },
      > @@ -946,7 +936,6 @@
      >           "MSRIndex": "0x1a6,0x1a7",
      >           "MSRValue": "0x3F803C0200",
      >           "Offcore": "1",
      > -        "PublicDescription": "Counts prefetch (that bring data
     to LLC only) code reads hit in the L3",
      >           "SampleAfterValue": "100003",
      >           "UMask": "0x1"
      >       },
      > @@ -959,7 +948,6 @@
      >           "MSRIndex": "0x1a6,0x1a7",
      >           "MSRValue": "0x3F803C0100",
      >           "Offcore": "1",
      > -        "PublicDescription": "Counts all prefetch (that bring
     data to LLC only) RFOs hit in the L3",
      >           "SampleAfterValue": "100003",
      >           "UMask": "0x1"
      >       },
      > @@ -973,4 +961,4 @@
      >           "SampleAfterValue": "100003",
      >           "UMask": "0x10"
      >       }
      > -]
      > \ No newline at end of file
      > +]
      > diff --git
     a/tools/perf/pmu-events/arch/x86/broadwellx/floating-point.json
     b/tools/perf/pmu-events/arch/x86/broadwellx/floating-point.json
      > index 9ad37dddb354..93bbc8600321 100644
      > --- a/tools/perf/pmu-events/arch/x86/broadwellx/floating-point.json
      > +++ b/tools/perf/pmu-events/arch/x86/broadwellx/floating-point.json
      > @@ -5,6 +5,7 @@
      >           "CounterHTOff": "0,1,2,3",
      >           "EventCode": "0xc7",
      >           "EventName": "FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE",
      > +        "PublicDescription": "Number of SSE/AVX computational
     128-bit packed double precision floating-point instructions retired;
     some instructions will count twice as noted below.  Each count
     represents 2 computation operations, one for each element.  Applies
     to SSE* and AVX* packed double precision floating-point
     instructions: ADD SUB HADD HSUB SUBADD MUL DIV MIN MAX SQRT DPP
     FM(N)ADD/SUB.  DPP and FM(N)ADD/SUB instructions count twice as they
     perform 2 calculations per element. The DAZ and FTZ flags in the
     MXCSR register need to be set when using these events.",
      >           "SampleAfterValue": "2000003",
      >           "UMask": "0x4"
      >       },
      > @@ -14,6 +15,7 @@
      >           "CounterHTOff": "0,1,2,3",
      >           "EventCode": "0xc7",
      >           "EventName": "FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE",
      > +        "PublicDescription": "Number of SSE/AVX computational
     128-bit packed single precision floating-point instructions retired;
     some instructions will count twice as noted below.  Each count
     represents 4 computation operations, one for each element.  Applies
     to SSE* and AVX* packed single precision floating-point
     instructions: ADD SUB HADD HSUB SUBADD MUL DIV MIN MAX SQRT RSQRT
     RCP DPP FM(N)ADD/SUB.  DPP and FM(N)ADD/SUB instructions count twice
     as they perform 2 calculations per element. The DAZ and FTZ flags in
     the MXCSR register need to be set when using these events.",
      >           "SampleAfterValue": "2000003",
      >           "UMask": "0x8"
      >       },
      > @@ -23,6 +25,7 @@
      >           "CounterHTOff": "0,1,2,3",
      >           "EventCode": "0xc7",
      >           "EventName": "FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE",
      > +        "PublicDescription": "Number of SSE/AVX computational
     256-bit packed double precision floating-point instructions retired;
     some instructions will count twice as noted below.  Each count
     represents 4 computation operations, one for each element.  Applies
     to SSE* and AVX* packed double precision floating-point
     instructions: ADD SUB HADD HSUB SUBADD MUL DIV MIN MAX SQRT
     FM(N)ADD/SUB.  FM(N)ADD/SUB instructions count twice as they perform
     2 calculations per element. The DAZ and FTZ flags in the MXCSR
     register need to be set when using these events.",
      >           "SampleAfterValue": "2000003",
      >           "UMask": "0x10"
      >       },
      > @@ -32,6 +35,7 @@
      >           "CounterHTOff": "0,1,2,3",
      >           "EventCode": "0xc7",
      >           "EventName": "FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE",
      > +        "PublicDescription": "Number of SSE/AVX computational
     256-bit packed single precision floating-point instructions retired;
     some instructions will count twice as noted below.  Each count
     represents 8 computation operations, one for each element.  Applies
     to SSE* and AVX* packed single precision floating-point
     instructions: ADD SUB HADD HSUB SUBADD MUL DIV MIN MAX SQRT RSQRT
     RCP DPP FM(N)ADD/SUB.  DPP and FM(N)ADD/SUB instructions count twice
     as they perform 2 calculations per element. The DAZ and FTZ flags in
     the MXCSR register need to be set when using these events.",
      >           "SampleAfterValue": "2000003",
      >           "UMask": "0x20"
      >       },
      > @@ -59,6 +63,7 @@
      >           "CounterHTOff": "0,1,2,3",
      >           "EventCode": "0xc7",
      >           "EventName": "FP_ARITH_INST_RETIRED.SCALAR",
      > +        "PublicDescription": "Number of SSE/AVX computational
     scalar single precision and double precision floating-point
     instructions retired; some instructions will count twice as noted
     below.  Each count represents 1 computational operation. Applies to
     SSE* and AVX* scalar single precision floating-point instructions:
     ADD SUB MUL DIV MIN MAX SQRT RSQRT RCP FM(N)ADD/SUB.  FM(N)ADD/SUB
     instructions count twice as they perform 2 calculations per element.
     The DAZ and FTZ flags in the MXCSR register need to be set when
     using these events.",
      >           "SampleAfterValue": "2000003",
      >           "UMask": "0x3"
      >       },
      > @@ -68,6 +73,7 @@
      >           "CounterHTOff": "0,1,2,3",
      >           "EventCode": "0xc7",
      >           "EventName": "FP_ARITH_INST_RETIRED.SCALAR_DOUBLE",
      > +        "PublicDescription": "Number of SSE/AVX computational
     scalar double precision floating-point instructions retired; some
     instructions will count twice as noted below.  Each count represents
     1 computational operation. Applies to SSE* and AVX* scalar double
     precision floating-point instructions: ADD SUB MUL DIV MIN MAX SQRT
     FM(N)ADD/SUB.  FM(N)ADD/SUB instructions count twice as they perform
     2 calculations per element. The DAZ and FTZ flags in the MXCSR
     register need to be set when using these events.",
      >           "SampleAfterValue": "2000003",
      >           "UMask": "0x1"
      >       },
      > @@ -77,6 +83,7 @@
      >           "CounterHTOff": "0,1,2,3",
      >           "EventCode": "0xc7",
      >           "EventName": "FP_ARITH_INST_RETIRED.SCALAR_SINGLE",
      > +        "PublicDescription": "Number of SSE/AVX computational
     scalar single precision floating-point instructions retired; some
     instructions will count twice as noted below.  Each count represents
     1 computational operation. Applies to SSE* and AVX* scalar single
     precision floating-point instructions: ADD SUB MUL DIV MIN MAX SQRT
     RSQRT RCP FM(N)ADD/SUB.  FM(N)ADD/SUB instructions count twice as
     they perform 2 calculations per element. The DAZ and FTZ flags in
     the MXCSR register need to be set when using these events.",
      >           "SampleAfterValue": "2000003",
      >           "UMask": "0x2"
      >       },
      > @@ -190,4 +197,4 @@
      >           "SampleAfterValue": "2000003",
      >           "UMask": "0x3"
      >       }
      > -]
      > \ No newline at end of file
      > +]
      > diff --git
     a/tools/perf/pmu-events/arch/x86/broadwellx/frontend.json
     b/tools/perf/pmu-events/arch/x86/broadwellx/frontend.json
      > index f0bcb945ff76..37ce8034b2ed 100644
      > --- a/tools/perf/pmu-events/arch/x86/broadwellx/frontend.json
      > +++ b/tools/perf/pmu-events/arch/x86/broadwellx/frontend.json
      > @@ -292,4 +292,4 @@
      >           "SampleAfterValue": "2000003",
      >           "UMask": "0x1"
      >       }
      > -]
      > \ No newline at end of file
      > +]
      > diff --git
     a/tools/perf/pmu-events/arch/x86/broadwellx/memory.json
     b/tools/perf/pmu-events/arch/x86/broadwellx/memory.json
      > index cce993b197e3..545f61f691b9 100644
      > --- a/tools/perf/pmu-events/arch/x86/broadwellx/memory.json
      > +++ b/tools/perf/pmu-events/arch/x86/broadwellx/memory.json
      > @@ -247,7 +247,6 @@
      >           "MSRIndex": "0x1a6,0x1a7",
      >           "MSRValue": "0x3FBFC00244",
      >           "Offcore": "1",
      > -        "PublicDescription": "Counts all demand & prefetch code
     reads miss in the L3",
      >           "SampleAfterValue": "100003",
      >           "UMask": "0x1"
      >       },
      > @@ -258,9 +257,8 @@
      >           "EventCode": "0xB7, 0xBB",
      >           "EventName":
     "OFFCORE_RESPONSE.ALL_CODE_RD.LLC_MISS.LOCAL_DRAM",
      >           "MSRIndex": "0x1a6,0x1a7",
      > -        "MSRValue": "0x0604000244",
      > +        "MSRValue": "0x604000244",
      >           "Offcore": "1",
      > -        "PublicDescription": "Counts all demand & prefetch code
     reads miss the L3 and the data is returned from local dram",
      >           "SampleAfterValue": "100003",
      >           "UMask": "0x1"
      >       },
      > @@ -273,7 +271,6 @@
      >           "MSRIndex": "0x1a6,0x1a7",
      >           "MSRValue": "0x3FBFC00091",
      >           "Offcore": "1",
      > -        "PublicDescription": "Counts all demand & prefetch data
     reads miss in the L3",
      >           "SampleAfterValue": "100003",
      >           "UMask": "0x1"
      >       },
      > @@ -284,9 +281,8 @@
      >           "EventCode": "0xB7, 0xBB",
      >           "EventName":
     "OFFCORE_RESPONSE.ALL_DATA_RD.LLC_MISS.LOCAL_DRAM",
      >           "MSRIndex": "0x1a6,0x1a7",
      > -        "MSRValue": "0x0604000091",
      > +        "MSRValue": "0x604000091",
      >           "Offcore": "1",
      > -        "PublicDescription": "Counts all demand & prefetch data
     reads miss the L3 and the data is returned from local dram",
      >           "SampleAfterValue": "100003",
      >           "UMask": "0x1"
      >       },
      > @@ -297,9 +293,8 @@
      >           "EventCode": "0xB7, 0xBB",
      >           "EventName":
     "OFFCORE_RESPONSE.ALL_DATA_RD.LLC_MISS.REMOTE_DRAM",
      >           "MSRIndex": "0x1a6,0x1a7",
      > -        "MSRValue": "0x063BC00091",
      > +        "MSRValue": "0x63BC00091",
      >           "Offcore": "1",
      > -        "PublicDescription": "Counts all demand & prefetch data
     reads miss the L3 and the data is returned from remote dram",
      >           "SampleAfterValue": "100003",
      >           "UMask": "0x1"
      >       },
      > @@ -312,7 +307,6 @@
      >           "MSRIndex": "0x1a6,0x1a7",
      >           "MSRValue": "0x103FC00091",
      >           "Offcore": "1",
      > -        "PublicDescription": "Counts all demand & prefetch data
     reads miss the L3 and the modified data is transferred from remote
     cache",
      >           "SampleAfterValue": "100003",
      >           "UMask": "0x1"
      >       },
      > @@ -323,9 +317,8 @@
      >           "EventCode": "0xB7, 0xBB",
      >           "EventName":
     "OFFCORE_RESPONSE.ALL_DATA_RD.LLC_MISS.REMOTE_HIT_FORWARD",
      >           "MSRIndex": "0x1a6,0x1a7",
      > -        "MSRValue": "0x087FC00091",
      > +        "MSRValue": "0x87FC00091",
      >           "Offcore": "1",
      > -        "PublicDescription": "Counts all demand & prefetch data
     reads miss the L3 and clean or shared data is transferred from
     remote cache",
      >           "SampleAfterValue": "100003",
      >           "UMask": "0x1"
      >       },
      > @@ -338,20 +331,18 @@
      >           "MSRIndex": "0x1a6,0x1a7",
      >           "MSRValue": "0x3FBFC007F7",
      >           "Offcore": "1",
      > -        "PublicDescription": "Counts all data/code/rfo reads
     (demand & prefetch) miss in the L3",
      >           "SampleAfterValue": "100003",
      >           "UMask": "0x1"
      >       },
      >       {
      > -        "BriefDescription": "Counts all data/code/rfo reads
     (demand & prefetch)miss the L3 and the data is returned from local
     dram",
      > +        "BriefDescription": "Counts all data/code/rfo reads
     (demand & prefetch) miss the L3 and the data is returned from local
     dram",
      >           "Counter": "0,1,2,3",
      >           "CounterHTOff": "0,1,2,3",
      >           "EventCode": "0xB7, 0xBB",
      >           "EventName":
     "OFFCORE_RESPONSE.ALL_READS.LLC_MISS.LOCAL_DRAM",
      >           "MSRIndex": "0x1a6,0x1a7",
      > -        "MSRValue": "0x06040007F7",
      > +        "MSRValue": "0x6040007F7",
      >           "Offcore": "1",
      > -        "PublicDescription": "Counts all data/code/rfo reads
     (demand & prefetch)miss the L3 and the data is returned from local
     dram",
      >           "SampleAfterValue": "100003",
      >           "UMask": "0x1"
      >       },
      > @@ -362,9 +353,8 @@
      >           "EventCode": "0xB7, 0xBB",
      >           "EventName":
     "OFFCORE_RESPONSE.ALL_READS.LLC_MISS.REMOTE_DRAM",
      >           "MSRIndex": "0x1a6,0x1a7",
      > -        "MSRValue": "0x063BC007F7",
      > +        "MSRValue": "0x63BC007F7",
      >           "Offcore": "1",
      > -        "PublicDescription": "Counts all data/code/rfo reads
     (demand & prefetch) miss the L3 and the data is returned from remote
     dram",
      >           "SampleAfterValue": "100003",
      >           "UMask": "0x1"
      >       },
      > @@ -377,7 +367,6 @@
      >           "MSRIndex": "0x1a6,0x1a7",
      >           "MSRValue": "0x103FC007F7",
      >           "Offcore": "1",
      > -        "PublicDescription": "Counts all data/code/rfo reads
     (demand & prefetch) miss the L3 and the modified data is transferred
     from remote cache",
      >           "SampleAfterValue": "100003",
      >           "UMask": "0x1"
      >       },
      > @@ -388,9 +377,8 @@
      >           "EventCode": "0xB7, 0xBB",
      >           "EventName":
     "OFFCORE_RESPONSE.ALL_READS.LLC_MISS.REMOTE_HIT_FORWARD",
      >           "MSRIndex": "0x1a6,0x1a7",
      > -        "MSRValue": "0x087FC007F7",
      > +        "MSRValue": "0x87FC007F7",
      >           "Offcore": "1",
      > -        "PublicDescription": "Counts all data/code/rfo reads
     (demand & prefetch) miss the L3 and clean or shared data is
     transferred from remote cache",
      >           "SampleAfterValue": "100003",
      >           "UMask": "0x1"
      >       },
      > @@ -403,7 +391,6 @@
      >           "MSRIndex": "0x1a6,0x1a7",
      >           "MSRValue": "0x3FBFC08FFF",
      >           "Offcore": "1",
      > -        "PublicDescription": "Counts all requests miss in the L3",
      >           "SampleAfterValue": "100003",
      >           "UMask": "0x1"
      >       },
      > @@ -416,7 +403,6 @@
      >           "MSRIndex": "0x1a6,0x1a7",
      >           "MSRValue": "0x3FBFC00122",
      >           "Offcore": "1",
      > -        "PublicDescription": "Counts all demand & prefetch RFOs
     miss in the L3",
      >           "SampleAfterValue": "100003",
      >           "UMask": "0x1"
      >       },
      > @@ -427,9 +413,8 @@
Quoted 9 lines
      >           "EventCode": "0xB7, 0xBB",
      >           "EventName":
     "OFFCORE_RESPONSE.ALL_RFO.LLC_MISS.LOCAL_DRAM",
      >           "MSRIndex": "0x1a6,0x1a7",
      > -        "MSRValue": "0x0604000122",
      > +        "MSRValue": "0x604000122",
      >           "Offcore": "1",
      > -        "PublicDescription": "Counts all demand & pref

--
Zhengjun Xing

Signature · 1 line
Zhengjun Xing
#8 lore raw

Re: [PATCH v1 02/31] perf vendor events: Update Intel broadwellx

Em Tue, Jul 26, 2022 at 01:19:02PM +0800, Xing Zhengjun escreveu:

Quoted 42 lines · Ian Rogers
Quoted 41 lines
Quoted 38 lines · Xing Zhengjun
Quoted 18 lines · Ian Rogers

On Sun, Jul 24, 2022, 6:34 PM Xing Zhengjun <[email protected] <mailto:[email protected]>> wrote:

Hi Ian,

On 7/23/2022 6:32 AM, Ian Rogers wrote:
> Use script at:
>
https://github.com/intel/event-converter-for-linux-perf/blob/master/download_and_gen.py
<https://github.com/intel/event-converter-for-linux-perf/blob/master/download_and_gen.py>
>

It is better to add the event JSON file version and TMA version for the future track. For example, the event list is based on broadwellx JSON file v19, the metrics are based on TMA4.4-full.

Quoted 7 lines

Thanks Xing,

I'll add that in v2. I'd skipped it this time as I'd been adding it to the mapfile. I'll also break apart the tremontx to snowridgex change to match yours. I also will rebase to see if that'll fix the git am issue. Apologies in advance to everyone's inboxes :-)

Hi Arnaldo,

Except for Snowridgex, I also post SPR/ADL/HSX/BDX event list last month, Can these be merged in or do I need to update them? Thanks.

https://lore.kernel.org/all/[email protected]/
https://lore.kernel.org/all/[email protected]/

https://lore.kernel.org/all/[email protected]/
https://lore.kernel.org/all/[email protected]/
https://lore.kernel.org/all/[email protected]/
https://lore.kernel.org/all/[email protected]/

Quoted 6 lines
Quoted 3 lines

Thanks Zhengjun,

I think those patches are all stale over what is posted here.

I think sometimes the middle version is also useful when we need back trace issues.

Yeah, I'll see if I can do the legwork on applying your previous patches, then Ian's so that we have the more finer grained history, which indeed can be useful.

- Arnaldo

Quoted 1,804 lines
Quoted 1,801 lines

Particular issues are:
- fixing the files to only contain ascii characters
- updating the version information in mapfile.csv
- for BDX and SPR (HSX isn't posted yet, although I've done some
private testing) adding in the metrics from:
https://github.com/intel/perfmon-metrics/tree/main/BDX/metrics/perf
https://github.com/intel/perfmon-metrics/tree/main/SPR/metrics/perf
- the event data on 01.org was updated. ADL, HSX and SPR were all
updated last Friday prior to my v1 patches.
- I also tested all the patches on the respective architectures,
hence discovering the fake event parsing fix that is in patch 1.

As stated in the cover letter, the goal is to make it so that what is in the Linux tree exactly matches the download_and_gen.py output, however, there were same changes I made that are in this pull request:
https://github.com/intel/event-converter-for-linux-perf/pull/15

Given the size of the new metrics and particularly the previously missed uncore data, it'd be nice to land the "compression" in:
https://lore.kernel.org/lkml/[email protected]/
which prior to this change was reducing the binary size for x86 by 12.5% as well as reducing over a megabytes worth of dirty pages for relocated data.

Thanks,
Ian

Quoted 1,774 lines
Quoted 1,762 lines

Thanks,
Ian

      > to download and generate the latest events and metrics. Manually copy
      > the broadwellx files into perf and update mapfile.csv.
      >
      > Tested with 'perf test':
      >   10: PMU events
          :
      >   10.1: PMU event table sanity
          : Ok
      >   10.2: PMU event map aliases
           : Ok
      >   10.3: Parsing of PMU event table metrics
          : Ok
      >   10.4: Parsing of PMU event table metrics with fake PMUs
           : Ok
      >   90: perf all metricgroups test
          : Ok
      >   91: perf all metrics test
           : Skip
      >   93: perf all PMU test
           : Ok
      >
      > Signed-off-by: Ian Rogers <[email protected]
     <mailto:[email protected]>>
      > ---
      >   .../arch/x86/broadwellx/bdx-metrics.json      |  570 ++-
      >   .../pmu-events/arch/x86/broadwellx/cache.json |   22 +-
      >   .../arch/x86/broadwellx/floating-point.json   |    9 +-
      >   .../arch/x86/broadwellx/frontend.json         |    2 +-
      >   .../arch/x86/broadwellx/memory.json           |   39 +-
      >   .../pmu-events/arch/x86/broadwellx/other.json |    2 +-
      >   .../arch/x86/broadwellx/pipeline.json         |    4 +-
      >   .../arch/x86/broadwellx/uncore-cache.json     | 3788
     ++++++++++++++++-
      >   .../x86/broadwellx/uncore-interconnect.json   | 1438 ++++++-
      >   .../arch/x86/broadwellx/uncore-memory.json    | 2849 ++++++++++++-
      >   .../arch/x86/broadwellx/uncore-other.json     | 3252 ++++++++++++++
      >   .../arch/x86/broadwellx/uncore-power.json     |  437 +-
      >   .../arch/x86/broadwellx/virtual-memory.json   |    2 +-
      >   tools/perf/pmu-events/arch/x86/mapfile.csv    |    2 +-
      >   14 files changed, 12103 insertions(+), 313 deletions(-)
      >   create mode 100644
     tools/perf/pmu-events/arch/x86/broadwellx/uncore-other.json
      >
      > diff --git
     a/tools/perf/pmu-events/arch/x86/broadwellx/bdx-metrics.json
     b/tools/perf/pmu-events/arch/x86/broadwellx/bdx-metrics.json
      > index b055947c0afe..720ee7c9332d 100644
      > --- a/tools/perf/pmu-events/arch/x86/broadwellx/bdx-metrics.json
      > +++ b/tools/perf/pmu-events/arch/x86/broadwellx/bdx-metrics.json
      > @@ -74,12 +74,6 @@
      >           "MetricGroup": "Branches;Fed;FetchBW",
      >           "MetricName": "UpTB"
      >       },
      > -    {
      > -        "BriefDescription": "Cycles Per Instruction (per Logical
     Processor)",
      > -        "MetricExpr": "1 / (INST_RETIRED.ANY /
     CPU_CLK_UNHALTED.THREAD)",
      > -        "MetricGroup": "Pipeline;Mem",
      > -        "MetricName": "CPI"
      > -    },
      >       {
      >           "BriefDescription": "Per-Logical Processor actual
     clocks when the Logical Processor is active.",
      >           "MetricExpr": "CPU_CLK_UNHALTED.THREAD",
      > @@ -130,43 +124,25 @@
      >           "MetricName": "FLOPc_SMT"
      >       },
      >       {
      > -        "BriefDescription": "Actual per-core usage of the
     Floating Point execution units (regardless of the vector width)",
      > +        "BriefDescription": "Actual per-core usage of the
     Floating Point non-X87 execution units (regardless of precision or
     vector-width)",
      >           "MetricExpr": "( (FP_ARITH_INST_RETIRED.SCALAR_SINGLE +
     FP_ARITH_INST_RETIRED.SCALAR_DOUBLE) +
     (FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE +
     FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE +
     FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE +
     FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE) ) / ( 2 *
     CPU_CLK_UNHALTED.THREAD )",
      >           "MetricGroup": "Cor;Flops;HPC",
      >           "MetricName": "FP_Arith_Utilization",
      > -        "PublicDescription": "Actual per-core usage of the
     Floating Point execution units (regardless of the vector width).
     Values > 1 are possible due to Fused-Multiply Add (FMA) counting."
      > +        "PublicDescription": "Actual per-core usage of the
     Floating Point non-X87 execution units (regardless of precision or
     vector-width). Values > 1 are possible due to ([BDW+] Fused-Multiply
     Add (FMA) counting - common; [ADL+] use all of ADD/MUL/FMA in Scalar
     or 128/256-bit vectors - less common)."
      >       },
      >       {
      > -        "BriefDescription": "Actual per-core usage of the
     Floating Point execution units (regardless of the vector width). SMT
     version; use when SMT is enabled and measuring per logical CPU.",
      > +        "BriefDescription": "Actual per-core usage of the
     Floating Point non-X87 execution units (regardless of precision or
     vector-width). SMT version; use when SMT is enabled and measuring
     per logical CPU.",
      >           "MetricExpr": "( (FP_ARITH_INST_RETIRED.SCALAR_SINGLE +
     FP_ARITH_INST_RETIRED.SCALAR_DOUBLE) +
     (FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE +
     FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE +
     FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE +
     FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE) ) / ( 2 * ( (
     CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 +
     CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ) )",
      >           "MetricGroup": "Cor;Flops;HPC_SMT",
      >           "MetricName": "FP_Arith_Utilization_SMT",
      > -        "PublicDescription": "Actual per-core usage of the
     Floating Point execution units (regardless of the vector width).
     Values > 1 are possible due to Fused-Multiply Add (FMA) counting.
     SMT version; use when SMT is enabled and measuring per logical CPU."
      > +        "PublicDescription": "Actual per-core usage of the
     Floating Point non-X87 execution units (regardless of precision or
     vector-width). Values > 1 are possible due to ([BDW+] Fused-Multiply
     Add (FMA) counting - common; [ADL+] use all of ADD/MUL/FMA in Scalar
     or 128/256-bit vectors - less common). SMT version; use when SMT is
     enabled and measuring per logical CPU."
      >       },
      >       {
      > -        "BriefDescription": "Instruction-Level-Parallelism
     (average number of uops executed when there is at least 1 uop
     executed)",
      > +        "BriefDescription": "Instruction-Level-Parallelism
     (average number of uops executed when there is execution) per-core",
      >           "MetricExpr": "UOPS_EXECUTED.THREAD / ((
     cpu@UOPS_EXECUTED.CORE\\,cmask\\=1@ / 2 ) if #SMT_on else
     UOPS_EXECUTED.CYCLES_GE_1_UOP_EXEC)",
      >           "MetricGroup": "Backend;Cor;Pipeline;PortsUtil",
      >           "MetricName": "ILP"
      >       },
      > -    {
      > -        "BriefDescription": "Branch Misprediction Cost: Fraction
     of TMA slots wasted per non-speculative branch misprediction
     (retired JEClear)",
      > -        "MetricExpr": " ( ((BR_MISP_RETIRED.ALL_BRANCHES / (
     BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT )) * ((
     UOPS_ISSUED.ANY - UOPS_RETIRED.RETIRE_SLOTS + 4 *
     INT_MISC.RECOVERY_CYCLES ) / (4 * CPU_CLK_UNHALTED.THREAD))) + (4 *
     IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 *
     CPU_CLK_UNHALTED.THREAD)) * (BR_MISP_RETIRED.ALL_BRANCHES * (12 * (
     BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT + BACLEARS.ANY )
     / CPU_CLK_UNHALTED.THREAD) / ( BR_MISP_RETIRED.ALL_BRANCHES +
     MACHINE_CLEARS.COUNT + BACLEARS.ANY )) / #(4 *
     IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 *
     CPU_CLK_UNHALTED.THREAD)) ) * (4 * CPU_CLK_UNHALTED.THREAD) /
     BR_MISP_RETIRED.ALL_BRANCHES",
      > -        "MetricGroup": "Bad;BrMispredicts",
      > -        "MetricName": "Branch_Misprediction_Cost"
      > -    },
      > -    {
      > -        "BriefDescription": "Branch Misprediction Cost: Fraction
     of TMA slots wasted per non-speculative branch misprediction
     (retired JEClear)",
      > -        "MetricExpr": " ( ((BR_MISP_RETIRED.ALL_BRANCHES / (
     BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT )) * ((
     UOPS_ISSUED.ANY - UOPS_RETIRED.RETIRE_SLOTS + 4 * (
     INT_MISC.RECOVERY_CYCLES_ANY / 2 ) ) / (4 * ( (
     CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 +
     CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK )
     )))) + (4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * (
     ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 +
     CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )))
     * (BR_MISP_RETIRED.ALL_BRANCHES * (12 * (
     BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT + BACLEARS.ANY )
     / CPU_CLK_UNHALTED.THREAD) / ( BR_MISP_RETIRED.ALL_BRANCHES +
     MACHINE_CLEARS.COUNT + BACLEARS.ANY )) / #(4 *
     IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * ( (
     CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 +
     CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )))
     ) * (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 +
     CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))
     / BR_MISP_RETIRED.ALL_BRANCHES",
      > -        "MetricGroup": "Bad;BrMispredicts_SMT",
      > -        "MetricName": "Branch_Misprediction_Cost_SMT"
      > -    },
      > -    {
      > -        "BriefDescription": "Number of Instructions per
     non-speculative Branch Misprediction (JEClear)",
      > -        "MetricExpr": "INST_RETIRED.ANY /
     BR_MISP_RETIRED.ALL_BRANCHES",
      > -        "MetricGroup": "Bad;BadSpec;BrMispredicts",
      > -        "MetricName": "IpMispredict"
      > -    },
      >       {
      >           "BriefDescription": "Core actual clocks when any
     Logical Processor is active on the Physical Core",
      >           "MetricExpr": "( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1
     + CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )",
      > @@ -256,6 +232,18 @@
      >           "MetricGroup": "Summary;TmaL1",
      >           "MetricName": "Instructions"
      >       },
      > +    {
      > +        "BriefDescription": "Average number of Uops retired in
     cycles where at least one uop has retired.",
      > +        "MetricExpr": "UOPS_RETIRED.RETIRE_SLOTS /
     cpu@UOPS_RETIRED.RETIRE_SLOTS\\,cmask\\=1@",
      > +        "MetricGroup": "Pipeline;Ret",
      > +        "MetricName": "Retire"
      > +    },
      > +    {
      > +        "BriefDescription": "",
      > +        "MetricExpr": "UOPS_EXECUTED.THREAD /
     cpu@UOPS_EXECUTED.THREAD\\,cmask\\=1@",
      > +        "MetricGroup": "Cor;Pipeline;PortsUtil;SMT",
      > +        "MetricName": "Execute"
      > +    },
      >       {
      >           "BriefDescription": "Fraction of Uops delivered by the
     DSB (aka Decoded ICache; or Uop Cache)",
      >           "MetricExpr": "IDQ.DSB_UOPS / (( IDQ.DSB_UOPS +
     LSD.UOPS + IDQ.MITE_UOPS + IDQ.MS_UOPS ) )",
      > @@ -263,35 +251,34 @@
      >           "MetricName": "DSB_Coverage"
      >       },
      >       {
      > -        "BriefDescription": "Actual Average Latency for L1
     data-cache miss demand load instructions (in core cycles)",
      > -        "MetricExpr": "L1D_PEND_MISS.PENDING / (
     MEM_LOAD_UOPS_RETIRED.L1_MISS + mem_load_uops_retired.hit_lfb )",
      > -        "MetricGroup": "Mem;MemoryBound;MemoryLat",
      > -        "MetricName": "Load_Miss_Real_Latency",
      > -        "PublicDescription": "Actual Average Latency for L1
     data-cache miss demand load instructions (in core cycles). Latency
     may be overestimated for multi-load instructions - e.g. repeat strings."
      > +        "BriefDescription": "Number of Instructions per
     non-speculative Branch Misprediction (JEClear) (lower number means
     higher occurrence rate)",
      > +        "MetricExpr": "INST_RETIRED.ANY /
     BR_MISP_RETIRED.ALL_BRANCHES",
      > +        "MetricGroup": "Bad;BadSpec;BrMispredicts",
      > +        "MetricName": "IpMispredict"
      >       },
      >       {
      > -        "BriefDescription": "Memory-Level-Parallelism (average
     number of L1 miss demand load when there is at least one such miss.
     Per-Logical Processor)",
      > -        "MetricExpr": "L1D_PEND_MISS.PENDING /
     L1D_PEND_MISS.PENDING_CYCLES",
      > -        "MetricGroup": "Mem;MemoryBound;MemoryBW",
      > -        "MetricName": "MLP"
      > +        "BriefDescription": "Branch Misprediction Cost: Fraction
     of TMA slots wasted per non-speculative branch misprediction
     (retired JEClear)",
      > +        "MetricExpr": " ( ((BR_MISP_RETIRED.ALL_BRANCHES / (
     BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT )) * ((
     UOPS_ISSUED.ANY - UOPS_RETIRED.RETIRE_SLOTS + 4 *
     INT_MISC.RECOVERY_CYCLES ) / (4 * CPU_CLK_UNHALTED.THREAD))) + (4 *
     IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 *
     CPU_CLK_UNHALTED.THREAD)) * (BR_MISP_RETIRED.ALL_BRANCHES * (12 * (
     BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT + BACLEARS.ANY )
     / CPU_CLK_UNHALTED.THREAD) / ( BR_MISP_RETIRED.ALL_BRANCHES +
     MACHINE_CLEARS.COUNT + BACLEARS.ANY )) / #(4 *
     IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 *
     CPU_CLK_UNHALTED.THREAD)) ) * (4 * CPU_CLK_UNHALTED.THREAD) /
     BR_MISP_RETIRED.ALL_BRANCHES",
      > +        "MetricGroup": "Bad;BrMispredicts",
      > +        "MetricName": "Branch_Misprediction_Cost"
      >       },
      >       {
      > -        "BriefDescription": "Average data fill bandwidth to the
     L1 data cache [GB / sec]",
      > -        "MetricExpr": "64 * L1D.REPLACEMENT / 1000000000 /
     duration_time",
      > -        "MetricGroup": "Mem;MemoryBW",
      > -        "MetricName": "L1D_Cache_Fill_BW"
      > +        "BriefDescription": "Branch Misprediction Cost: Fraction
     of TMA slots wasted per non-speculative branch misprediction
     (retired JEClear)",
      > +        "MetricExpr": " ( ((BR_MISP_RETIRED.ALL_BRANCHES / (
     BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT )) * ((
     UOPS_ISSUED.ANY - UOPS_RETIRED.RETIRE_SLOTS + 4 * (
     INT_MISC.RECOVERY_CYCLES_ANY / 2 ) ) / (4 * ( (
     CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 +
     CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK )
     )))) + (4 * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * (
     ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 +
     CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )))
     * (BR_MISP_RETIRED.ALL_BRANCHES * (12 * (
     BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT + BACLEARS.ANY )
     / CPU_CLK_UNHALTED.THREAD) / ( BR_MISP_RETIRED.ALL_BRANCHES +
     MACHINE_CLEARS.COUNT + BACLEARS.ANY )) / #(4 *
     IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / (4 * ( (
     CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 +
     CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) )))
     ) * (4 * ( ( CPU_CLK_UNHALTED.THREAD / 2 ) * ( 1 +
     CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE / CPU_CLK_UNHALTED.REF_XCLK ) ))
     / BR_MISP_RETIRED.ALL_BRANCHES",
      > +        "MetricGroup": "Bad;BrMispredicts_SMT",
      > +        "MetricName": "Branch_Misprediction_Cost_SMT"
      >       },
      >       {
      > -        "BriefDescription": "Average data fill bandwidth to the
     L2 cache [GB / sec]",
      > -        "MetricExpr": "64 * L2_LINES_IN.ALL / 1000000000 /
     duration_time",
      > -        "MetricGroup": "Mem;MemoryBW",
      > -        "MetricName": "L2_Cache_Fill_BW"
      > +        "BriefDescription": "Actual Average Latency for L1
     data-cache miss demand load operations (in core cycles)",
      > +        "MetricExpr": "L1D_PEND_MISS.PENDING / (
     MEM_LOAD_UOPS_RETIRED.L1_MISS + mem_load_uops_retired.hit_lfb )",
      > +        "MetricGroup": "Mem;MemoryBound;MemoryLat",
      > +        "MetricName": "Load_Miss_Real_Latency"
      >       },
      >       {
      > -        "BriefDescription": "Average per-core data fill
     bandwidth to the L3 cache [GB / sec]",
      > -        "MetricExpr": "64 * LONGEST_LAT_CACHE.MISS / 1000000000
     / duration_time",
      > -        "MetricGroup": "Mem;MemoryBW",
      > -        "MetricName": "L3_Cache_Fill_BW"
      > +        "BriefDescription": "Memory-Level-Parallelism (average
     number of L1 miss demand load when there is at least one such miss.
     Per-Logical Processor)",
      > +        "MetricExpr": "L1D_PEND_MISS.PENDING /
     L1D_PEND_MISS.PENDING_CYCLES",
      > +        "MetricGroup": "Mem;MemoryBound;MemoryBW",
      > +        "MetricName": "MLP"
      >       },
      >       {
      >           "BriefDescription": "L1 cache true misses per kilo
     instruction for retired demand loads",
      > @@ -306,13 +293,13 @@
      >           "MetricName": "L2MPKI"
      >       },
      >       {
      > -        "BriefDescription": "L2 cache misses per kilo
     instruction for all request types (including speculative)",
      > +        "BriefDescription": "L2 cache ([RKL+] true) misses per
     kilo instruction for all request types (including speculative)",
      >           "MetricExpr": "1000 * L2_RQSTS.MISS / INST_RETIRED.ANY",
      >           "MetricGroup": "Mem;CacheMisses;Offcore",
      >           "MetricName": "L2MPKI_All"
      >       },
      >       {
      > -        "BriefDescription": "L2 cache misses per kilo
     instruction for all demand loads  (including speculative)",
      > +        "BriefDescription": "L2 cache ([RKL+] true) misses per
     kilo instruction for all demand loads  (including speculative)",
      >           "MetricExpr": "1000 * L2_RQSTS.DEMAND_DATA_RD_MISS /
     INST_RETIRED.ANY",
      >           "MetricGroup": "Mem;CacheMisses",
      >           "MetricName": "L2MPKI_Load"
      > @@ -348,6 +335,48 @@
      >           "MetricGroup": "Mem;MemoryTLB_SMT",
      >           "MetricName": "Page_Walks_Utilization_SMT"
      >       },
      > +    {
      > +        "BriefDescription": "Average per-core data fill
     bandwidth to the L1 data cache [GB / sec]",
      > +        "MetricExpr": "64 * L1D.REPLACEMENT / 1000000000 /
     duration_time",
      > +        "MetricGroup": "Mem;MemoryBW",
      > +        "MetricName": "L1D_Cache_Fill_BW"
      > +    },
      > +    {
      > +        "BriefDescription": "Average per-core data fill
     bandwidth to the L2 cache [GB / sec]",
      > +        "MetricExpr": "64 * L2_LINES_IN.ALL / 1000000000 /
     duration_time",
      > +        "MetricGroup": "Mem;MemoryBW",
      > +        "MetricName": "L2_Cache_Fill_BW"
      > +    },
      > +    {
      > +        "BriefDescription": "Average per-core data fill
     bandwidth to the L3 cache [GB / sec]",
      > +        "MetricExpr": "64 * LONGEST_LAT_CACHE.MISS / 1000000000
     / duration_time",
      > +        "MetricGroup": "Mem;MemoryBW",
      > +        "MetricName": "L3_Cache_Fill_BW"
      > +    },
      > +    {
      > +        "BriefDescription": "Average per-thread data fill
     bandwidth to the L1 data cache [GB / sec]",
      > +        "MetricExpr": "(64 * L1D.REPLACEMENT / 1000000000 /
     duration_time)",
      > +        "MetricGroup": "Mem;MemoryBW",
      > +        "MetricName": "L1D_Cache_Fill_BW_1T"
      > +    },
      > +    {
      > +        "BriefDescription": "Average per-thread data fill
     bandwidth to the L2 cache [GB / sec]",
      > +        "MetricExpr": "(64 * L2_LINES_IN.ALL / 1000000000 /
     duration_time)",
      > +        "MetricGroup": "Mem;MemoryBW",
      > +        "MetricName": "L2_Cache_Fill_BW_1T"
      > +    },
      > +    {
      > +        "BriefDescription": "Average per-thread data fill
     bandwidth to the L3 cache [GB / sec]",
      > +        "MetricExpr": "(64 * LONGEST_LAT_CACHE.MISS / 1000000000
     / duration_time)",
      > +        "MetricGroup": "Mem;MemoryBW",
      > +        "MetricName": "L3_Cache_Fill_BW_1T"
      > +    },
      > +    {
      > +        "BriefDescription": "Average per-thread data access
     bandwidth to the L3 cache [GB / sec]",
      > +        "MetricExpr": "0",
      > +        "MetricGroup": "Mem;MemoryBW;Offcore",
      > +        "MetricName": "L3_Cache_Access_BW_1T"
      > +    },
      >       {
      >           "BriefDescription": "Average CPU Utilization",
      >           "MetricExpr": "CPU_CLK_UNHALTED.REF_TSC / msr@tsc@",
      > @@ -364,7 +393,8 @@
      >           "BriefDescription": "Giga Floating Point Operations Per
     Second",
      >           "MetricExpr": "( ( 1 * (
     FP_ARITH_INST_RETIRED.SCALAR_SINGLE +
     FP_ARITH_INST_RETIRED.SCALAR_DOUBLE ) + 2 *
     FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE + 4 * (
     FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE +
     FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE ) + 8 *
     FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE ) / 1000000000 ) /
     duration_time",
      >           "MetricGroup": "Cor;Flops;HPC",
      > -        "MetricName": "GFLOPs"
      > +        "MetricName": "GFLOPs",
      > +        "PublicDescription": "Giga Floating Point Operations Per
     Second. Aggregate across all supported options of: FP precisions,
     scalar and vector instructions, vector-width and AMX engine."
      >       },
      >       {
      >           "BriefDescription": "Average Frequency Utilization
     relative nominal frequency",
      > @@ -461,5 +491,439 @@
      >           "MetricExpr": "(cstate_pkg@c7\\-residency@ / msr@tsc@)
     * 100",
      >           "MetricGroup": "Power",
      >           "MetricName": "C7_Pkg_Residency"
      > +    },
      > +    {
      > +        "BriefDescription": "CPU operating frequency (in GHz)",
      > +        "MetricExpr": "( CPU_CLK_UNHALTED.THREAD /
     CPU_CLK_UNHALTED.REF_TSC * #SYSTEM_TSC_FREQ ) / 1000000000",
      > +        "MetricGroup": "",
      > +        "MetricName": "cpu_operating_frequency",
      > +        "ScaleUnit": "1GHz"
      > +    },
      > +    {
      > +        "BriefDescription": "Cycles per instruction retired;
     indicating how much time each executed instruction took; in units of
     cycles.",
      > +        "MetricExpr": "CPU_CLK_UNHALTED.THREAD / INST_RETIRED.ANY",
      > +        "MetricGroup": "",
      > +        "MetricName": "cpi",
      > +        "ScaleUnit": "1per_instr"
      > +    },
      > +    {
      > +        "BriefDescription": "The ratio of number of completed
     memory load instructions to the total number completed instructions",
      > +        "MetricExpr": "MEM_UOPS_RETIRED.ALL_LOADS /
     INST_RETIRED.ANY",
      > +        "MetricGroup": "",
      > +        "MetricName": "loads_per_instr",
      > +        "ScaleUnit": "1per_instr"
      > +    },
      > +    {
      > +        "BriefDescription": "The ratio of number of completed
     memory store instructions to the total number completed instructions",
      > +        "MetricExpr": "MEM_UOPS_RETIRED.ALL_STORES /
     INST_RETIRED.ANY",
      > +        "MetricGroup": "",
      > +        "MetricName": "stores_per_instr",
      > +        "ScaleUnit": "1per_instr"
      > +    },
      > +    {
      > +        "BriefDescription": "Ratio of number of requests missing
     L1 data cache (includes data+rfo w/ prefetches) to the total number
     of completed instructions",
      > +        "MetricExpr": "L1D.REPLACEMENT / INST_RETIRED.ANY",
      > +        "MetricGroup": "",
      > +        "MetricName":
     "l1d_mpi_includes_data_plus_rfo_with_prefetches",
      > +        "ScaleUnit": "1per_instr"
      > +    },
      > +    {
      > +        "BriefDescription": "Ratio of number of demand load
     requests hitting in L1 data cache to the total number of completed
     instructions",
      > +        "MetricExpr": "MEM_LOAD_UOPS_RETIRED.L1_HIT /
     INST_RETIRED.ANY",
      > +        "MetricGroup": "",
      > +        "MetricName": "l1d_demand_data_read_hits_per_instr",
      > +        "ScaleUnit": "1per_instr"
      > +    },
      > +    {
      > +        "BriefDescription": "Ratio of number of code read
     requests missing in L1 instruction cache (includes prefetches) to
     the total number of completed instructions",
      > +        "MetricExpr": "L2_RQSTS.ALL_CODE_RD / INST_RETIRED.ANY",
      > +        "MetricGroup": "",
      > +        "MetricName":
     "l1_i_code_read_misses_with_prefetches_per_instr",
      > +        "ScaleUnit": "1per_instr"
      > +    },
      > +    {
      > +        "BriefDescription": "Ratio of number of completed demand
     load requests hitting in L2 cache to the total number of completed
     instructions",
      > +        "MetricExpr": "MEM_LOAD_UOPS_RETIRED.L2_HIT /
     INST_RETIRED.ANY",
      > +        "MetricGroup": "",
      > +        "MetricName": "l2_demand_data_read_hits_per_instr",
      > +        "ScaleUnit": "1per_instr"
      > +    },
      > +    {
      > +        "BriefDescription": "Ratio of number of requests missing
     L2 cache (includes code+data+rfo w/ prefetches) to the total number
     of completed instructions",
      > +        "MetricExpr": "L2_LINES_IN.ALL / INST_RETIRED.ANY",
      > +        "MetricGroup": "",
      > +        "MetricName":
     "l2_mpi_includes_code_plus_data_plus_rfo_with_prefetches",
      > +        "ScaleUnit": "1per_instr"
      > +    },
      > +    {
      > +        "BriefDescription": "Ratio of number of completed data
     read request missing L2 cache to the total number of completed
     instructions",
      > +        "MetricExpr": "MEM_LOAD_UOPS_RETIRED.L2_MISS /
     INST_RETIRED.ANY",
      > +        "MetricGroup": "",
      > +        "MetricName": "l2_demand_data_read_mpi",
      > +        "ScaleUnit": "1per_instr"
      > +    },
      > +    {
      > +        "BriefDescription": "Ratio of number of code read
     request missing L2 cache to the total number of completed instructions",
      > +        "MetricExpr": "L2_RQSTS.CODE_RD_MISS / INST_RETIRED.ANY",
      > +        "MetricGroup": "",
      > +        "MetricName": "l2_demand_code_mpi",
      > +        "ScaleUnit": "1per_instr"
      > +    },
      > +    {
      > +        "BriefDescription": "Ratio of number of data read
     requests missing last level core cache (includes demand w/
     prefetches) to the total number of completed instructions",
      > +        "MetricExpr": "(
     cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x182@ +
     cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x192@ ) /
     INST_RETIRED.ANY",
      > +        "MetricGroup": "",
      > +        "MetricName": "llc_data_read_mpi_demand_plus_prefetch",
      > +        "ScaleUnit": "1per_instr"
      > +    },
      > +    {
      > +        "BriefDescription": "Ratio of number of code read
     requests missing last level core cache (includes demand w/
     prefetches) to the total number of completed instructions",
      > +        "MetricExpr": "(
     cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x181@ +
     cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x191@ ) /
     INST_RETIRED.ANY",
      > +        "MetricGroup": "",
      > +        "MetricName": "llc_code_read_mpi_demand_plus_prefetch",
      > +        "ScaleUnit": "1per_instr"
      > +    },
      > +    {
      > +        "BriefDescription": "Average latency of a last level
     cache (LLC) demand and prefetch data read miss (read memory access)
     in nano seconds",
      > +        "MetricExpr": "( 1000000000 * (
     cbox@UNC_C_TOR_OCCUPANCY.MISS_OPCODE\\,filter_opc\\=0x182@ /
     cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x182@ ) / (
     UNC_C_CLOCKTICKS / ( source_count(UNC_C_CLOCKTICKS) * #num_packages
     ) ) ) * duration_time",
      > +        "MetricGroup": "",
      > +        "MetricName":
     "llc_data_read_demand_plus_prefetch_miss_latency",
      > +        "ScaleUnit": "1ns"
      > +    },
      > +    {
      > +        "BriefDescription": "Average latency of a last level
     cache (LLC) demand and prefetch data read miss (read memory access)
     addressed to local memory in nano seconds",
      > +        "MetricExpr": "( 1000000000 * (
     cbox@UNC_C_TOR_OCCUPANCY.MISS_LOCAL_OPCODE\\,filter_opc\\=0x182@ /
     cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x182@ ) / (
     UNC_C_CLOCKTICKS / ( source_count(UNC_C_CLOCKTICKS) * #num_packages
     ) ) ) * duration_time",
      > +        "MetricGroup": "",
      > +        "MetricName":
     "llc_data_read_demand_plus_prefetch_miss_latency_for_local_requests",
      > +        "ScaleUnit": "1ns"
      > +    },
      > +    {
      > +        "BriefDescription": "Average latency of a last level
     cache (LLC) demand and prefetch data read miss (read memory access)
     addressed to remote memory in nano seconds",
      > +        "MetricExpr": "( 1000000000 * (
     cbox@UNC_C_TOR_OCCUPANCY.MISS_REMOTE_OPCODE\\,filter_opc\\=0x182@ /
     cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x182@ ) / (
     UNC_C_CLOCKTICKS / ( source_count(UNC_C_CLOCKTICKS) * #num_packages
     ) ) ) * duration_time",
      > +        "MetricGroup": "",
      > +        "MetricName":
     "llc_data_read_demand_plus_prefetch_miss_latency_for_remote_requests",
      > +        "ScaleUnit": "1ns"
      > +    },
      > +    {
      > +        "BriefDescription": "Ratio of number of completed page
     walks (for all page sizes) caused by a code fetch to the total
     number of completed instructions. This implies it missed in the ITLB
     (Instruction TLB) and further levels of TLB.",
      > +        "MetricExpr": "ITLB_MISSES.WALK_COMPLETED /
     INST_RETIRED.ANY",
      > +        "MetricGroup": "",
      > +        "MetricName": "itlb_mpi",
      > +        "ScaleUnit": "1per_instr"
      > +    },
      > +    {
      > +        "BriefDescription": "Ratio of number of completed page
     walks (for 2 megabyte and 4 megabyte page sizes) caused by a code
     fetch to the total number of completed instructions. This implies it
     missed in the Instruction Translation Lookaside Buffer (ITLB) and
     further levels of TLB.",
      > +        "MetricExpr": "ITLB_MISSES.WALK_COMPLETED_2M_4M /
     INST_RETIRED.ANY",
      > +        "MetricGroup": "",
      > +        "MetricName": "itlb_large_page_mpi",
      > +        "ScaleUnit": "1per_instr"
      > +    },
      > +    {
      > +        "BriefDescription": "Ratio of number of completed page
     walks (for all page sizes) caused by demand data loads to the total
     number of completed instructions. This implies it missed in the DTLB
     and further levels of TLB.",
      > +        "MetricExpr": "DTLB_LOAD_MISSES.WALK_COMPLETED /
     INST_RETIRED.ANY",
      > +        "MetricGroup": "",
      > +        "MetricName": "dtlb_load_mpi",
      > +        "ScaleUnit": "1per_instr"
      > +    },
      > +    {
      > +        "BriefDescription": "Ratio of number of completed page
     walks (for all page sizes) caused by demand data stores to the total
     number of completed instructions. This implies it missed in the DTLB
     and further levels of TLB.",
      > +        "MetricExpr": "DTLB_STORE_MISSES.WALK_COMPLETED /
     INST_RETIRED.ANY",
      > +        "MetricGroup": "",
      > +        "MetricName": "dtlb_store_mpi",
      > +        "ScaleUnit": "1per_instr"
      > +    },
      > +    {
      > +        "BriefDescription": "Memory read that miss the last
     level cache (LLC) addressed to local DRAM as a percentage of total
     memory read accesses, does not include LLC prefetches.",
      > +        "MetricExpr": "100 *
     cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x182@ / (
     cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x182@ +
     cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x182@ )",
      > +        "MetricGroup": "",
      > +        "MetricName": "numa_percent_reads_addressed_to_local_dram",
      > +        "ScaleUnit": "1%"
      > +    },
      > +    {
      > +        "BriefDescription": "Memory reads that miss the last
     level cache (LLC) addressed to remote DRAM as a percentage of total
     memory read accesses, does not include LLC prefetches.",
      > +        "MetricExpr": "100 *
     cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x182@ / (
     cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x182@ +
     cbox@UNC_C_TOR_INSERTS.MISS_OPCODE\\,filter_opc\\=0x182@ )",
      > +        "MetricGroup": "",
      > +        "MetricName": "numa_percent_reads_addressed_to_remote_dram",
      > +        "ScaleUnit": "1%"
      > +    },
      > +    {
      > +        "BriefDescription": "Uncore operating frequency in GHz",
      > +        "MetricExpr": "UNC_C_CLOCKTICKS / (
     source_count(UNC_C_CLOCKTICKS) * #num_packages ) / 1000000000",
      > +        "MetricGroup": "",
      > +        "MetricName": "uncore_frequency",
      > +        "ScaleUnit": "1GHz"
      > +    },
      > +    {
      > +        "BriefDescription": "Intel(R) Quick Path Interconnect
     (QPI) data transmit bandwidth (MB/sec)",
      > +        "MetricExpr": "( UNC_Q_TxL_FLITS_G0.DATA * 8 / 1000000)
     / duration_time",
      > +        "MetricGroup": "",
      > +        "MetricName": "qpi_data_transmit_bw_only_data",
      > +        "ScaleUnit": "1MB/s"
      > +    },
      > +    {
      > +        "BriefDescription": "DDR memory read bandwidth (MB/sec)",
      > +        "MetricExpr": "( UNC_M_CAS_COUNT.RD * 64 / 1000000) /
     duration_time",
      > +        "MetricGroup": "",
      > +        "MetricName": "memory_bandwidth_read",
      > +        "ScaleUnit": "1MB/s"
      > +    },
      > +    {
      > +        "BriefDescription": "DDR memory write bandwidth (MB/sec)",
      > +        "MetricExpr": "( UNC_M_CAS_COUNT.WR * 64 / 1000000) /
     duration_time",
      > +        "MetricGroup": "",
      > +        "MetricName": "memory_bandwidth_write",
      > +        "ScaleUnit": "1MB/s"
      > +    },
      > +    {
      > +        "BriefDescription": "DDR memory bandwidth (MB/sec)",
      > +        "MetricExpr": "(( UNC_M_CAS_COUNT.RD +
     UNC_M_CAS_COUNT.WR ) * 64 / 1000000) / duration_time",
      > +        "MetricGroup": "",
      > +        "MetricName": "memory_bandwidth_total",
      > +        "ScaleUnit": "1MB/s"
      > +    },
      > +    {
      > +        "BriefDescription": "Bandwidth of IO reads that are
     initiated by end device controllers that are requesting memory from
     the CPU.",
      > +        "MetricExpr": "(
     cbox@UNC_C_TOR_INSERTS.OPCODE\\,filter_opc\\=0x19e@ * 64 / 1000000)
     / duration_time",
      > +        "MetricGroup": "",
      > +        "MetricName": "io_bandwidth_read",
      > +        "ScaleUnit": "1MB/s"
      > +    },
      > +    {
      > +        "BriefDescription": "Bandwidth of IO writes that are
     initiated by end device controllers that are writing memory to the
     CPU.",
      > +        "MetricExpr": "((
     cbox@UNC_C_TOR_INSERTS.OPCODE\\,filter_opc\\=0x1c8\\,filter_tid\\=0x3e@
     +
     cbox@UNC_C_TOR_INSERTS.OPCODE\\,filter_opc\\=0x180\\,filter_tid\\=0x3e@
     ) * 64 / 1000000) / duration_time",
      > +        "MetricGroup": "",
      > +        "MetricName": "io_bandwidth_write",
      > +        "ScaleUnit": "1MB/s"
      > +    },
      > +    {
      > +        "BriefDescription": "Uops delivered from decoded
     instruction cache (decoded stream buffer or DSB) as a percent of
     total uops delivered to Instruction Decode Queue",
      > +        "MetricExpr": "100 * ( IDQ.DSB_UOPS / UOPS_ISSUED.ANY )",
      > +        "MetricGroup": "",
      > +        "MetricName":
     "percent_uops_delivered_from_decoded_icache_dsb",
      > +        "ScaleUnit": "1%"
      > +    },
      > +    {
      > +        "BriefDescription": "Uops delivered from legacy decode
     pipeline (Micro-instruction Translation Engine or MITE) as a percent
     of total uops delivered to Instruction Decode Queue",
      > +        "MetricExpr": "100 * ( IDQ.MITE_UOPS / UOPS_ISSUED.ANY )",
      > +        "MetricGroup": "",
      > +        "MetricName":
     "percent_uops_delivered_from_legacy_decode_pipeline_mite",
      > +        "ScaleUnit": "1%"
      > +    },
      > +    {
      > +        "BriefDescription": "Uops delivered from microcode
     sequencer (MS) as a percent of total uops delivered to Instruction
     Decode Queue",
      > +        "MetricExpr": "100 * ( IDQ.MS_UOPS / UOPS_ISSUED.ANY )",
      > +        "MetricGroup": "",
      > +        "MetricName":
     "percent_uops_delivered_from_microcode_sequencer_ms",
      > +        "ScaleUnit": "1%"
      > +    },
      > +    {
      > +        "BriefDescription": "Uops delivered from loop stream
     detector(LSD) as a percent of total uops delivered to Instruction
     Decode Queue",
      > +        "MetricExpr": "100 * ( LSD.UOPS / UOPS_ISSUED.ANY )",
      > +        "MetricGroup": "",
      > +        "MetricName":
     "percent_uops_delivered_from_loop_stream_detector_lsd",
      > +        "ScaleUnit": "1%"
      > +    },
      > +    {
      > +        "BriefDescription": "This category represents fraction
     of slots where the processor's Frontend undersupplies its Backend.
     Frontend denotes the first part of the processor core responsible to
     fetch operations that are executed later on by the Backend part.
     Within the Frontend; a branch predictor predicts the next address to
     fetch; cache-lines are fetched from the memory subsystem; parsed
     into instructions; and lastly decoded into micro-operations (uops).
     Ideally the Frontend can issue Machine_Width uops every cycle to the
     Backend. Frontend Bound denotes unutilized issue-slots when there is
     no Backend stall; i.e. bubbles where Frontend delivered no uops
     while Backend could have accepted them. For example; stalls due to
     instruction-cache misses would be categorized under Frontend Bound.",
      > +        "MetricExpr": "100 * ( IDQ_UOPS_NOT_DELIVERED.CORE / ( (
     4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else (
     CPU_CLK_UNHALTED.THREAD ) ) ) )",
      > +        "MetricGroup": "TmaL1;PGO",
      > +        "MetricName": "tma_frontend_bound_percent",
      > +        "ScaleUnit": "1%"
      > +    },
      > +    {
      > +        "BriefDescription": "This metric represents fraction of
     slots the CPU was stalled due to Frontend latency issues.  For
     example; instruction-cache misses; iTLB misses or fetch stalls after
     a branch misprediction are categorized under Frontend Latency. In
     such cases; the Frontend eventually delivers no uops for some period.",
      > +        "MetricExpr": "100 * ( ( 4 ) *
     IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / ( ( 4 ) * ( (
     CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else (
     CPU_CLK_UNHALTED.THREAD ) ) ) )",
      > +        "MetricGroup":
     "Frontend;TmaL2;m_tma_frontend_bound_percent",
      > +        "MetricName": "tma_fetch_latency_percent",
      > +        "ScaleUnit": "1%"
      > +    },
      > +    {
      > +        "BriefDescription": "This metric represents fraction of
     cycles the CPU was stalled due to instruction cache misses.",
      > +        "MetricExpr": "100 * ( ICACHE.IFDATA_STALL / (
     CPU_CLK_UNHALTED.THREAD ) )",
      > +        "MetricGroup":
     "BigFoot;FetchLat;IcMiss;TmaL3;m_tma_fetch_latency_percent",
      > +        "MetricName": "tma_icache_misses_percent",
      > +        "ScaleUnit": "1%"
      > +    },
      > +    {
      > +        "BriefDescription": "This metric represents fraction of
     cycles the CPU was stalled due to Instruction TLB (ITLB) misses.",
      > +        "MetricExpr": "100 * ( ( 14 * ITLB_MISSES.STLB_HIT +
     cpu@ITLB_MISSES.WALK_DURATION\\,cmask\\=0x1@ + 7 *
     ITLB_MISSES.WALK_COMPLETED ) / ( CPU_CLK_UNHALTED.THREAD ) )",
      > +        "MetricGroup":
     "BigFoot;FetchLat;MemoryTLB;TmaL3;m_tma_fetch_latency_percent",
      > +        "MetricName": "tma_itlb_misses_percent",
      > +        "ScaleUnit": "1%"
      > +    },
      > +    {
      > +        "BriefDescription": "This metric represents fraction of
     cycles the CPU was stalled due to Branch Resteers. Branch Resteers
     estimates the Frontend delay in fetching operations from corrected
     path; following all sorts of miss-predicted branches. For example;
     branchy code with lots of miss-predictions might get categorized
     under Branch Resteers. Note the value of this node may overlap with
     its siblings.",
      > +        "MetricExpr": "100 * ( ( 12 ) * (
     BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT + BACLEARS.ANY )
     / ( CPU_CLK_UNHALTED.THREAD ) )",
      > +        "MetricGroup": "FetchLat;TmaL3;m_tma_fetch_latency_percent",
      > +        "MetricName": "tma_branch_resteers_percent",
      > +        "ScaleUnit": "1%"
      > +    },
      > +    {
      > +        "BriefDescription": "This metric represents fraction of
     cycles the CPU was stalled due to switches from DSB to MITE
     pipelines. The DSB (decoded i-cache) is a Uop Cache where the
     front-end directly delivers Uops (micro operations) avoiding heavy
     x86 decoding. The DSB pipeline has shorter latency and delivered
     higher bandwidth than the MITE (legacy instruction decode pipeline).
     Switching between the two pipelines can cause penalties hence this
     metric measures the exposed penalty.",
      > +        "MetricExpr": "100 * ( DSB2MITE_SWITCHES.PENALTY_CYCLES
     / ( CPU_CLK_UNHALTED.THREAD ) )",
      > +        "MetricGroup":
     "DSBmiss;FetchLat;TmaL3;m_tma_fetch_latency_percent",
      > +        "MetricName": "tma_dsb_switches_percent",
      > +        "ScaleUnit": "1%"
      > +    },
      > +    {
      > +        "BriefDescription": "This metric represents fraction of
     cycles CPU was stalled due to Length Changing Prefixes (LCPs). Using
     proper compiler flags or Intel Compiler by default will certainly
     avoid this. #Link: Optimization Guide about LCP BKMs.",
      > +        "MetricExpr": "100 * ( ILD_STALL.LCP / (
     CPU_CLK_UNHALTED.THREAD ) )",
      > +        "MetricGroup": "FetchLat;TmaL3;m_tma_fetch_latency_percent",
      > +        "MetricName": "tma_lcp_percent",
      > +        "ScaleUnit": "1%"
      > +    },
      > +    {
      > +        "BriefDescription": "This metric estimates the fraction
     of cycles when the CPU was stalled due to switches of uop delivery
     to the Microcode Sequencer (MS). Commonly used instructions are
     optimized for delivery by the DSB (decoded i-cache) or MITE (legacy
     instruction decode) pipelines. Certain operations cannot be handled
     natively by the execution pipeline; and must be performed by
     microcode (small programs injected into the execution stream).
     Switching to the MS too often can negatively impact performance. The
     MS is designated to deliver long uop flows required by CISC
     instructions like CPUID; or uncommon conditions like Floating Point
     Assists when dealing with Denormals.",
      > +        "MetricExpr": "100 * ( ( 2 ) * IDQ.MS_SWITCHES / (
     CPU_CLK_UNHALTED.THREAD ) )",
      > +        "MetricGroup":
     "FetchLat;MicroSeq;TmaL3;m_tma_fetch_latency_percent",
      > +        "MetricName": "tma_ms_switches_percent",
      > +        "ScaleUnit": "1%"
      > +    },
      > +    {
      > +        "BriefDescription": "This metric represents fraction of
     slots the CPU was stalled due to Frontend bandwidth issues.  For
     example; inefficiencies at the instruction decoders; or restrictions
     for caching in the DSB (decoded uops cache) are categorized under
     Fetch Bandwidth. In such cases; the Frontend typically delivers
     suboptimal amount of uops to the Backend.",
      > +        "MetricExpr": "100 * ( ( IDQ_UOPS_NOT_DELIVERED.CORE / (
     ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else (
     CPU_CLK_UNHALTED.THREAD ) ) ) ) - ( ( 4 ) *
     IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / ( ( 4 ) * ( (
     CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else (
     CPU_CLK_UNHALTED.THREAD ) ) ) ) )",
      > +        "MetricGroup":
     "FetchBW;Frontend;TmaL2;m_tma_frontend_bound_percent",
      > +        "MetricName": "tma_fetch_bandwidth_percent",
      > +        "ScaleUnit": "1%"
      > +    },
      > +    {
      > +        "BriefDescription": "This metric represents Core
     fraction of cycles in which CPU was likely limited due to the MITE
     pipeline (the legacy decode pipeline). This pipeline is used for
     code that was not pre-cached in the DSB or LSD. For example;
     inefficiencies due to asymmetric decoders; use of long immediate or
     LCP can manifest as MITE fetch bandwidth bottleneck.",
      > +        "MetricExpr": "100 * ( ( IDQ.ALL_MITE_CYCLES_ANY_UOPS -
     IDQ.ALL_MITE_CYCLES_4_UOPS ) / ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 )
     if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) / 2 )",
      > +        "MetricGroup":
     "DSBmiss;FetchBW;TmaL3;m_tma_fetch_bandwidth_percent",
      > +        "MetricName": "tma_mite_percent",
      > +        "ScaleUnit": "1%"
      > +    },
      > +    {
      > +        "BriefDescription": "This metric represents Core
     fraction of cycles in which CPU was likely limited due to DSB
     (decoded uop cache) fetch pipeline.  For example; inefficient
     utilization of the DSB cache structure or bank conflict when reading
     from it; are categorized here.",
      > +        "MetricExpr": "100 * ( ( IDQ.ALL_DSB_CYCLES_ANY_UOPS -
     IDQ.ALL_DSB_CYCLES_4_UOPS ) / ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 )
     if #SMT_on else ( CPU_CLK_UNHALTED.THREAD ) ) / 2 )",
      > +        "MetricGroup":
     "DSB;FetchBW;TmaL3;m_tma_fetch_bandwidth_percent",
      > +        "MetricName": "tma_dsb_percent",
      > +        "ScaleUnit": "1%"
      > +    },
      > +    {
      > +        "BriefDescription": "This category represents fraction
     of slots wasted due to incorrect speculations. This include slots
     used to issue uops that do not eventually get retired and slots for
     which the issue-pipeline was blocked due to recovery from earlier
     incorrect speculation. For example; wasted work due to
     miss-predicted branches are categorized under Bad Speculation
     category. Incorrect data speculation followed by Memory Ordering
     Nukes is another example.",
      > +        "MetricExpr": "100 * ( ( UOPS_ISSUED.ANY - (
     UOPS_RETIRED.RETIRE_SLOTS ) + ( 4 ) * ( (
     INT_MISC.RECOVERY_CYCLES_ANY / 2 ) if #SMT_on else
     INT_MISC.RECOVERY_CYCLES ) ) / ( ( 4 ) * ( (
     CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else (
     CPU_CLK_UNHALTED.THREAD ) ) ) )",
      > +        "MetricGroup": "TmaL1",
      > +        "MetricName": "tma_bad_speculation_percent",
      > +        "ScaleUnit": "1%"
      > +    },
      > +    {
      > +        "BriefDescription": "This metric represents fraction of
     slots the CPU has wasted due to Branch Misprediction.  These slots
     are either wasted by uops fetched from an incorrectly speculated
     program path; or stalls when the out-of-order part of the machine
     needs to recover its state from a speculative path.",
      > +        "MetricExpr": "100 * ( ( BR_MISP_RETIRED.ALL_BRANCHES /
     ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT ) ) * ( (
     UOPS_ISSUED.ANY - ( UOPS_RETIRED.RETIRE_SLOTS ) + ( 4 ) * ( (
     INT_MISC.RECOVERY_CYCLES_ANY / 2 ) if #SMT_on else
     INT_MISC.RECOVERY_CYCLES ) ) / ( ( 4 ) * ( (
     CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else (
     CPU_CLK_UNHALTED.THREAD ) ) ) ) )",
      > +        "MetricGroup":
     "BadSpec;BrMispredicts;TmaL2;m_tma_bad_speculation_percent",
      > +        "MetricName": "tma_branch_mispredicts_percent",
      > +        "ScaleUnit": "1%"
      > +    },
      > +    {
      > +        "BriefDescription": "This metric represents fraction of
     slots the CPU has wasted due to Machine Clears.  These slots are
     either wasted by uops fetched prior to the clear; or stalls the
     out-of-order portion of the machine needs to recover its state after
     the clear. For example; this can happen due to memory ordering Nukes
     (e.g. Memory Disambiguation) or Self-Modifying-Code (SMC) nukes.",
      > +        "MetricExpr": "100 * ( ( ( UOPS_ISSUED.ANY - (
     UOPS_RETIRED.RETIRE_SLOTS ) + ( 4 ) * ( (
     INT_MISC.RECOVERY_CYCLES_ANY / 2 ) if #SMT_on else
     INT_MISC.RECOVERY_CYCLES ) ) / ( ( 4 ) * ( (
     CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else (
     CPU_CLK_UNHALTED.THREAD ) ) ) ) - ( ( BR_MISP_RETIRED.ALL_BRANCHES /
     ( BR_MISP_RETIRED.ALL_BRANCHES + MACHINE_CLEARS.COUNT ) ) * ( (
     UOPS_ISSUED.ANY - ( UOPS_RETIRED.RETIRE_SLOTS ) + ( 4 ) * ( (
     INT_MISC.RECOVERY_CYCLES_ANY / 2 ) if #SMT_on else
     INT_MISC.RECOVERY_CYCLES ) ) / ( ( 4 ) * ( (
     CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else (
     CPU_CLK_UNHALTED.THREAD ) ) ) ) ) )",
      > +        "MetricGroup":
     "BadSpec;MachineClears;TmaL2;m_tma_bad_speculation_percent",
      > +        "MetricName": "tma_machine_clears_percent",
      > +        "ScaleUnit": "1%"
      > +    },
      > +    {
      > +        "BriefDescription": "This category represents fraction
     of slots where no uops are being delivered due to a lack of required
     resources for accepting new uops in the Backend. Backend is the
     portion of the processor core where the out-of-order scheduler
     dispatches ready uops into their respective execution units; and
     once completed these uops get retired according to program order.
     For example; stalls due to data-cache misses or stalls due to the
     divider unit being overloaded are both categorized under Backend
     Bound. Backend Bound is further divided into two main categories:
     Memory Bound and Core Bound.",
      > +        "MetricExpr": "100 * ( 1 - ( (
     IDQ_UOPS_NOT_DELIVERED.CORE / ( ( 4 ) * ( (
     CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else (
     CPU_CLK_UNHALTED.THREAD ) ) ) ) + ( ( UOPS_ISSUED.ANY - (
     UOPS_RETIRED.RETIRE_SLOTS ) + ( 4 ) * ( (
     INT_MISC.RECOVERY_CYCLES_ANY / 2 ) if #SMT_on else
     INT_MISC.RECOVERY_CYCLES ) ) / ( ( 4 ) * ( (
     CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else (
     CPU_CLK_UNHALTED.THREAD ) ) ) ) + ( ( UOPS_RETIRED.RETIRE_SLOTS ) /
     ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else (
     CPU_CLK_UNHALTED.THREAD ) ) ) ) ) )",
      > +        "MetricGroup": "TmaL1",
      > +        "MetricName": "tma_backend_bound_percent",
      > +        "ScaleUnit": "1%"
      > +    },
      > +    {
      > +        "BriefDescription": "This metric represents fraction of
     slots the Memory subsystem within the Backend was a bottleneck.
     Memory Bound estimates fraction of slots where pipeline is likely
     stalled due to demand load or store instructions. This accounts
     mainly for (1) non-completed in-flight memory demand loads which
     coincides with execution units starvation; in addition to (2) cases
     where stores could impose backpressure on the pipeline when many of
     them get buffered at the same time (less common out of the two).",
      > +        "MetricExpr": "100 * ( ( ( CYCLE_ACTIVITY.STALLS_MEM_ANY
     + RESOURCE_STALLS.SB <http://RESOURCE_STALLS.SB> ) / ( (
     CYCLE_ACTIVITY.STALLS_TOTAL + UOPS_EXECUTED.CYCLES_GE_1_UOP_EXEC - (
     UOPS_EXECUTED.CYCLES_GE_3_UOPS_EXEC if ( ( INST_RETIRED.ANY / (
     CPU_CLK_UNHALTED.THREAD ) ) > 1.8 ) else
     UOPS_EXECUTED.CYCLES_GE_2_UOPS_EXEC ) - ( RS_EVENTS.EMPTY_CYCLES if
     ( ( ( 4 ) * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / ( ( 4
     ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else (
     CPU_CLK_UNHALTED.THREAD ) ) ) ) > 0.1 ) else 0 ) +
     RESOURCE_STALLS.SB <http://RESOURCE_STALLS.SB> ) ) ) * ( 1 - ( (
     IDQ_UOPS_NOT_DELIVERED.CORE / ( ( 4 ) * ( (
     CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else (
     CPU_CLK_UNHALTED.THREAD ) ) ) ) + ( ( UOPS_ISSUED.ANY - (
     UOPS_RETIRED.RETIRE_SLOTS ) + ( 4 ) * ( (
     INT_MISC.RECOVERY_CYCLES_ANY / 2 ) if #SMT_on else
     INT_MISC.RECOVERY_CYCLES ) ) / ( ( 4 ) * ( (
     CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else (
     CPU_CLK_UNHALTED.THREAD ) ) ) ) + ( ( UOPS_RETIRED.RETIRE_SLOTS ) /
     ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else (
     CPU_CLK_UNHALTED.THREAD ) ) ) ) ) ) )",
      > +        "MetricGroup": "Backend;TmaL2;m_tma_backend_bound_percent",
      > +        "MetricName": "tma_memory_bound_percent",
      > +        "ScaleUnit": "1%"
      > +    },
      > +    {
      > +        "BriefDescription": "This metric estimates how often the
     CPU was stalled without loads missing the L1 data cache.  The L1
     data cache typically has the shortest latency.  However; in certain
     cases like loads blocked on older stores; a load might suffer due to
     high latency even though it is being satisfied by the L1. Another
     example is loads who miss in the TLB. These cases are characterized
     by execution unit stalls; while some non-completed demand load lives
     in the machine without having that demand load missing the L1 cache.",
      > +        "MetricExpr": "100 * ( max( (
     CYCLE_ACTIVITY.STALLS_MEM_ANY - CYCLE_ACTIVITY.STALLS_L1D_MISS ) / (
     CPU_CLK_UNHALTED.THREAD ) , 0 ) )",
      > +        "MetricGroup":
     "CacheMisses;MemoryBound;TmaL3mem;TmaL3;m_tma_memory_bound_percent",
      > +        "MetricName": "tma_l1_bound_percent",
      > +        "ScaleUnit": "1%"
      > +    },
      > +    {
      > +        "BriefDescription": "This metric estimates how often the
     CPU was stalled due to L2 cache accesses by loads.  Avoiding cache
     misses (i.e. L1 misses/L2 hits) can improve the latency and increase
     performance.",
      > +        "MetricExpr": "100 * ( ( CYCLE_ACTIVITY.STALLS_L1D_MISS
     - CYCLE_ACTIVITY.STALLS_L2_MISS ) / ( CPU_CLK_UNHALTED.THREAD ) )",
      > +        "MetricGroup":
     "CacheMisses;MemoryBound;TmaL3mem;TmaL3;m_tma_memory_bound_percent",
      > +        "MetricName": "tma_l2_bound_percent",
      > +        "ScaleUnit": "1%"
      > +    },
      > +    {
      > +        "BriefDescription": "This metric estimates how often the
     CPU was stalled due to loads accesses to L3 cache or contended with
     a sibling Core.  Avoiding cache misses (i.e. L2 misses/L3 hits) can
     improve the latency and increase performance.",
      > +        "MetricExpr": "100 * ( ( MEM_LOAD_UOPS_RETIRED.L3_HIT /
     ( MEM_LOAD_UOPS_RETIRED.L3_HIT + ( 7 ) *
     MEM_LOAD_UOPS_RETIRED.L3_MISS ) ) * CYCLE_ACTIVITY.STALLS_L2_MISS /
     ( CPU_CLK_UNHALTED.THREAD ) )",
      > +        "MetricGroup":
     "CacheMisses;MemoryBound;TmaL3mem;TmaL3;m_tma_memory_bound_percent",
      > +        "MetricName": "tma_l3_bound_percent",
      > +        "ScaleUnit": "1%"
      > +    },
      > +    {
      > +        "BriefDescription": "This metric estimates how often the
     CPU was stalled on accesses to external memory (DRAM) by loads.
     Better caching can improve the latency and increase performance.",
      > +        "MetricExpr": "100 * ( min( ( ( 1 - (
     MEM_LOAD_UOPS_RETIRED.L3_HIT / ( MEM_LOAD_UOPS_RETIRED.L3_HIT + ( 7
     ) * MEM_LOAD_UOPS_RETIRED.L3_MISS ) ) ) *
     CYCLE_ACTIVITY.STALLS_L2_MISS / ( CPU_CLK_UNHALTED.THREAD ) ) , ( 1
     ) ) )",
      > +        "MetricGroup":
     "MemoryBound;TmaL3mem;TmaL3;m_tma_memory_bound_percent",
      > +        "MetricName": "tma_dram_bound_percent",
      > +        "ScaleUnit": "1%"
      > +    },
      > +    {
      > +        "BriefDescription": "This metric estimates how often CPU
     was stalled  due to RFO store memory accesses; RFO store issue a
     read-for-ownership request before the write. Even though store
     accesses do not typically stall out-of-order CPUs; there are few
     cases where stores can lead to actual stalls. This metric will be
     flagged should RFO stores be a bottleneck.",
      > +        "MetricExpr": "100 * ( RESOURCE_STALLS.SB
     <http://RESOURCE_STALLS.SB> / ( CPU_CLK_UNHALTED.THREAD ) )",
      > +        "MetricGroup":
     "MemoryBound;TmaL3mem;TmaL3;m_tma_memory_bound_percent",
      > +        "MetricName": "tma_store_bound_percent",
      > +        "ScaleUnit": "1%"
      > +    },
      > +    {
      > +        "BriefDescription": "This metric represents fraction of
     slots where Core non-memory issues were of a bottleneck.  Shortage
     in hardware compute resources; or dependencies in software's
     instructions are both categorized under Core Bound. Hence it may
     indicate the machine ran out of an out-of-order resource; certain
     execution units are overloaded or dependencies in program's data- or
     instruction-flow are limiting the performance (e.g. FP-chained
     long-latency arithmetic operations).",
      > +        "MetricExpr": "100 * ( ( 1 - ( (
     IDQ_UOPS_NOT_DELIVERED.CORE / ( ( 4 ) * ( (
     CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else (
     CPU_CLK_UNHALTED.THREAD ) ) ) ) + ( ( UOPS_ISSUED.ANY - (
     UOPS_RETIRED.RETIRE_SLOTS ) + ( 4 ) * ( (
     INT_MISC.RECOVERY_CYCLES_ANY / 2 ) if #SMT_on else
     INT_MISC.RECOVERY_CYCLES ) ) / ( ( 4 ) * ( (
     CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else (
     CPU_CLK_UNHALTED.THREAD ) ) ) ) + ( ( UOPS_RETIRED.RETIRE_SLOTS ) /
     ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else (
     CPU_CLK_UNHALTED.THREAD ) ) ) ) ) ) - ( ( (
     CYCLE_ACTIVITY.STALLS_MEM_ANY + RESOURCE_STALLS.SB
     <http://RESOURCE_STALLS.SB> ) / ( ( CYCLE_ACTIVITY.STALLS_TOTAL +
     UOPS_EXECUTED.CYCLES_GE_1_UOP_EXEC - (
     UOPS_EXECUTED.CYCLES_GE_3_UOPS_EXEC if ( ( INST_RETIRED.ANY / (
     CPU_CLK_UNHALTED.THREAD ) ) > 1.8 ) else
     UOPS_EXECUTED.CYCLES_GE_2_UOPS_EXEC ) - ( RS_EVENTS.EMPTY_CYCLES if
     ( ( ( 4 ) * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / ( ( 4
     ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else (
     CPU_CLK_UNHALTED.THREAD ) ) ) ) > 0.1 ) else 0 ) +
     RESOURCE_STALLS.SB <http://RESOURCE_STALLS.SB> ) ) ) * ( 1 - ( (
     IDQ_UOPS_NOT_DELIVERED.CORE / ( ( 4 ) * ( (
     CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else (
     CPU_CLK_UNHALTED.THREAD ) ) ) ) + ( ( UOPS_ISSUED.ANY - (
     UOPS_RETIRED.RETIRE_SLOTS ) + ( 4 ) * ( (
     INT_MISC.RECOVERY_CYCLES_ANY / 2 ) if #SMT_on else
     INT_MISC.RECOVERY_CYCLES ) ) / ( ( 4 ) * ( (
     CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else (
     CPU_CLK_UNHALTED.THREAD ) ) ) ) + ( ( UOPS_RETIRED.RETIRE_SLOTS ) /
     ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else (
     CPU_CLK_UNHALTED.THREAD ) ) ) ) ) ) ) )",
      > +        "MetricGroup":
     "Backend;TmaL2;Compute;m_tma_backend_bound_percent",
      > +        "MetricName": "tma_core_bound_percent",
      > +        "ScaleUnit": "1%"
      > +    },
      > +    {
      > +        "BriefDescription": "This metric represents fraction of
     cycles where the Divider unit was active. Divide and square root
     instructions are performed by the Divider unit and can take
     considerably longer latency than integer or Floating Point addition;
     subtraction; or multiplication.",
      > +        "MetricExpr": "100 * ( ARITH.FPU_DIV_ACTIVE / ( (
     CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else (
     CPU_CLK_UNHALTED.THREAD ) ) )",
      > +        "MetricGroup": "TmaL3;m_tma_core_bound_percent",
      > +        "MetricName": "tma_divider_percent",
      > +        "ScaleUnit": "1%"
      > +    },
      > +    {
      > +        "BriefDescription": "This metric estimates fraction of
     cycles the CPU performance was potentially limited due to Core
     computation issues (non divider-related).  Two distinct categories
     can be attributed into this metric: (1) heavy data-dependency among
     contiguous instructions would manifest in this metric - such cases
     are often referred to as low Instruction Level Parallelism (ILP).
     (2) Contention on some hardware execution unit other than Divider.
     For example; when there are too many multiply operations.",
      > +        "MetricExpr": "100 * ( ( ( ( CYCLE_ACTIVITY.STALLS_TOTAL
     + UOPS_EXECUTED.CYCLES_GE_1_UOP_EXEC - (
     UOPS_EXECUTED.CYCLES_GE_3_UOPS_EXEC if ( ( INST_RETIRED.ANY / (
     CPU_CLK_UNHALTED.THREAD ) ) > 1.8 ) else
     UOPS_EXECUTED.CYCLES_GE_2_UOPS_EXEC ) - ( RS_EVENTS.EMPTY_CYCLES if
     ( ( ( 4 ) * IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE / ( ( 4
     ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else (
     CPU_CLK_UNHALTED.THREAD ) ) ) ) > 0.1 ) else 0 ) +
     RESOURCE_STALLS.SB <http://RESOURCE_STALLS.SB> ) ) -
     RESOURCE_STALLS.SB <http://RESOURCE_STALLS.SB> -
     CYCLE_ACTIVITY.STALLS_MEM_ANY ) / ( CPU_CLK_UNHALTED.THREAD ) )",
      > +        "MetricGroup": "PortsUtil;TmaL3;m_tma_core_bound_percent",
      > +        "MetricName": "tma_ports_utilization_percent",
      > +        "ScaleUnit": "1%"
      > +    },
      > +    {
      > +        "BriefDescription": "This category represents fraction
     of slots utilized by useful work i.e. issued uops that eventually
     get retired. Ideally; all pipeline slots would be attributed to the
     Retiring category.  Retiring of 100% would indicate the maximum
     Pipeline_Width throughput was achieved.  Maximizing Retiring
     typically increases the Instructions-per-cycle (see IPC metric).
     Note that a high Retiring value does not necessary mean there is no
     room for more performance.  For example; Heavy-operations or
     Microcode Assists are categorized under Retiring. They often
     indicate suboptimal performance and can often be optimized or
     avoided. ",
      > +        "MetricExpr": "100 * ( ( UOPS_RETIRED.RETIRE_SLOTS ) / (
     ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else (
     CPU_CLK_UNHALTED.THREAD ) ) ) )",
      > +        "MetricGroup": "TmaL1",
      > +        "MetricName": "tma_retiring_percent",
      > +        "ScaleUnit": "1%"
      > +    },
      > +    {
      > +        "BriefDescription": "This metric represents fraction of
     slots where the CPU was retiring light-weight operations --
     instructions that require no more than one uop (micro-operation).
     This correlates with total number of instructions used by the
     program. A uops-per-instruction (see UPI metric) ratio of 1 or less
     should be expected for decently optimized software running on Intel
     Core/Xeon products. While this often indicates efficient X86
     instructions were executed; high value does not necessarily mean
     better performance cannot be achieved.",
      > +        "MetricExpr": "100 * ( ( ( UOPS_RETIRED.RETIRE_SLOTS ) /
     ( ( 4 ) * ( ( CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else (
     CPU_CLK_UNHALTED.THREAD ) ) ) ) - ( ( ( ( UOPS_RETIRED.RETIRE_SLOTS
     ) / UOPS_ISSUED.ANY ) * IDQ.MS_UOPS / ( ( 4 ) * ( (
     CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else (
     CPU_CLK_UNHALTED.THREAD ) ) ) ) ) )",
      > +        "MetricGroup": "Retire;TmaL2;m_tma_retiring_percent",
      > +        "MetricName": "tma_light_operations_percent",
      > +        "ScaleUnit": "1%"
      > +    },
      > +    {
      > +        "BriefDescription": "This metric represents overall
     arithmetic floating-point (FP) operations fraction the CPU has
     executed (retired). Note this metric's value may exceed its parent
     due to use of \"Uops\" CountDomain and FMA double-counting.",
      > +        "MetricExpr": "100 * ( ( INST_RETIRED.X87 * ( (
     UOPS_RETIRED.RETIRE_SLOTS ) / INST_RETIRED.ANY ) / (
     UOPS_RETIRED.RETIRE_SLOTS ) ) + ( (
     FP_ARITH_INST_RETIRED.SCALAR_SINGLE +
     FP_ARITH_INST_RETIRED.SCALAR_DOUBLE ) / ( UOPS_RETIRED.RETIRE_SLOTS
     ) ) + ( min( ( ( FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE +
     FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE +
     FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE +
     FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE ) / (
     UOPS_RETIRED.RETIRE_SLOTS ) ) , ( 1 ) ) ) )",
      > +        "MetricGroup": "HPC;TmaL3;m_tma_light_operations_percent",
      > +        "MetricName": "tma_fp_arith_percent",
      > +        "ScaleUnit": "1%"
      > +    },
      > +    {
      > +        "BriefDescription": "This metric represents fraction of
     slots where the CPU was retiring heavy-weight operations --
     instructions that require two or more uops or microcoded sequences.
     This highly-correlates with the uop length of these
     instructions/sequences.",
      > +        "MetricExpr": "100 * ( ( ( ( UOPS_RETIRED.RETIRE_SLOTS )
     / UOPS_ISSUED.ANY ) * IDQ.MS_UOPS / ( ( 4 ) * ( (
     CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else (
     CPU_CLK_UNHALTED.THREAD ) ) ) ) )",
      > +        "MetricGroup": "Retire;TmaL2;m_tma_retiring_percent",
      > +        "MetricName": "tma_heavy_operations_percent",
      > +        "ScaleUnit": "1%"
      > +    },
      > +    {
      > +        "BriefDescription": "This metric represents fraction of
     slots the CPU was retiring uops fetched by the Microcode Sequencer
     (MS) unit.  The MS is used for CISC instructions not supported by
     the default decoders (like repeat move strings; or CPUID); or by
     microcode assists used to address some operation modes (like in
     Floating Point assists). These cases can often be avoided.",
      > +        "MetricExpr": "100 * ( ( ( UOPS_RETIRED.RETIRE_SLOTS ) /
     UOPS_ISSUED.ANY ) * IDQ.MS_UOPS / ( ( 4 ) * ( (
     CPU_CLK_UNHALTED.THREAD_ANY / 2 ) if #SMT_on else (
     CPU_CLK_UNHALTED.THREAD ) ) ) )",
      > +        "MetricGroup":
     "MicroSeq;TmaL3;m_tma_heavy_operations_percent",
      > +        "MetricName": "tma_microcode_sequencer_percent",
      > +        "ScaleUnit": "1%"
      >       }
      >   ]
      > diff --git a/tools/perf/pmu-events/arch/x86/broadwellx/cache.json
     b/tools/perf/pmu-events/arch/x86/broadwellx/cache.json
      > index 127abe08362f..2efc4c0ee740 100644
      > --- a/tools/perf/pmu-events/arch/x86/broadwellx/cache.json
      > +++ b/tools/perf/pmu-events/arch/x86/broadwellx/cache.json
      > @@ -814,9 +814,8 @@
      >           "EventCode": "0xB7, 0xBB",
      >           "EventName":
     "OFFCORE_RESPONSE.ALL_CODE_RD.LLC_HIT.HIT_OTHER_CORE_NO_FWD",
      >           "MSRIndex": "0x1a6,0x1a7",
      > -        "MSRValue": "0x04003C0244",
      > +        "MSRValue": "0x4003C0244",
      >           "Offcore": "1",
      > -        "PublicDescription": "Counts all demand & prefetch code
     reads hit in the L3 and the snoops to sibling cores hit in either
     E/S state and the line is not forwarded",
      >           "SampleAfterValue": "100003",
      >           "UMask": "0x1"
      >       },
      > @@ -829,7 +828,6 @@
      >           "MSRIndex": "0x1a6,0x1a7",
      >           "MSRValue": "0x10003C0091",
      >           "Offcore": "1",
      > -        "PublicDescription": "Counts all demand & prefetch data
     reads hit in the L3 and the snoop to one of the sibling cores hits
     the line in M state and the line is forwarded",
      >           "SampleAfterValue": "100003",
      >           "UMask": "0x1"
      >       },
      > @@ -840,9 +838,8 @@
      >           "EventCode": "0xB7, 0xBB",
      >           "EventName":
     "OFFCORE_RESPONSE.ALL_DATA_RD.LLC_HIT.HIT_OTHER_CORE_NO_FWD",
      >           "MSRIndex": "0x1a6,0x1a7",
      > -        "MSRValue": "0x04003C0091",
      > +        "MSRValue": "0x4003C0091",
      >           "Offcore": "1",
      > -        "PublicDescription": "Counts all demand & prefetch data
     reads hit in the L3 and the snoops to sibling cores hit in either
     E/S state and the line is not forwarded",
      >           "SampleAfterValue": "100003",
      >           "UMask": "0x1"
      >       },
      > @@ -855,7 +852,6 @@
      >           "MSRIndex": "0x1a6,0x1a7",
      >           "MSRValue": "0x10003C07F7",
      >           "Offcore": "1",
      > -        "PublicDescription": "Counts all data/code/rfo reads
     (demand & prefetch) hit in the L3 and the snoop to one of the
     sibling cores hits the line in M state and the line is forwarded",
      >           "SampleAfterValue": "100003",
      >           "UMask": "0x1"
      >       },
      > @@ -866,9 +862,8 @@
      >           "EventCode": "0xB7, 0xBB",
      >           "EventName":
     "OFFCORE_RESPONSE.ALL_READS.LLC_HIT.HIT_OTHER_CORE_NO_FWD",
      >           "MSRIndex": "0x1a6,0x1a7",
      > -        "MSRValue": "0x04003C07F7",
      > +        "MSRValue": "0x4003C07F7",
      >           "Offcore": "1",
      > -        "PublicDescription": "Counts all data/code/rfo reads
     (demand & prefetch) hit in the L3 and the snoops to sibling cores
     hit in either E/S state and the line is not forwarded",
      >           "SampleAfterValue": "100003",
      >           "UMask": "0x1"
      >       },
      > @@ -881,7 +876,6 @@
      >           "MSRIndex": "0x1a6,0x1a7",
      >           "MSRValue": "0x3F803C8FFF",
      >           "Offcore": "1",
      > -        "PublicDescription": "Counts all requests hit in the L3",
      >           "SampleAfterValue": "100003",
      >           "UMask": "0x1"
      >       },
      > @@ -894,7 +888,6 @@
      >           "MSRIndex": "0x1a6,0x1a7",
      >           "MSRValue": "0x10003C0122",
      >           "Offcore": "1",
      > -        "PublicDescription": "Counts all demand & prefetch RFOs
     hit in the L3 and the snoop to one of the sibling cores hits the
     line in M state and the line is forwarded",
      >           "SampleAfterValue": "100003",
      >           "UMask": "0x1"
      >       },
      > @@ -905,9 +898,8 @@
      >           "EventCode": "0xB7, 0xBB",
      >           "EventName":
     "OFFCORE_RESPONSE.ALL_RFO.LLC_HIT.HIT_OTHER_CORE_NO_FWD",
      >           "MSRIndex": "0x1a6,0x1a7",
      > -        "MSRValue": "0x04003C0122",
      > +        "MSRValue": "0x4003C0122",
      >           "Offcore": "1",
      > -        "PublicDescription": "Counts all demand & prefetch RFOs
     hit in the L3 and the snoops to sibling cores hit in either E/S
     state and the line is not forwarded",
      >           "SampleAfterValue": "100003",
      >           "UMask": "0x1"
      >       },
      > @@ -920,7 +912,6 @@
      >           "MSRIndex": "0x1a6,0x1a7",
      >           "MSRValue": "0x3F803C0002",
      >           "Offcore": "1",
      > -        "PublicDescription": "Counts all demand data writes
     (RFOs) hit in the L3",
      >           "SampleAfterValue": "100003",
      >           "UMask": "0x1"
      >       },
      > @@ -933,7 +924,6 @@
      >           "MSRIndex": "0x1a6,0x1a7",
      >           "MSRValue": "0x10003C0002",
      >           "Offcore": "1",
      > -        "PublicDescription": "Counts all demand data writes
     (RFOs) hit in the L3 and the snoop to one of the sibling cores hits
     the line in M state and the line is forwarded",
      >           "SampleAfterValue": "100003",
      >           "UMask": "0x1"
      >       },
      > @@ -946,7 +936,6 @@
      >           "MSRIndex": "0x1a6,0x1a7",
      >           "MSRValue": "0x3F803C0200",
      >           "Offcore": "1",
      > -        "PublicDescription": "Counts prefetch (that bring data
     to LLC only) code reads hit in the L3",
      >           "SampleAfterValue": "100003",
      >           "UMask": "0x1"
      >       },
      > @@ -959,7 +948,6 @@
      >           "MSRIndex": "0x1a6,0x1a7",
      >           "MSRValue": "0x3F803C0100",
      >           "Offcore": "1",
      > -        "PublicDescription": "Counts all prefetch (that bring
     data to LLC only) RFOs hit in the L3",
      >           "SampleAfterValue": "100003",
      >           "UMask": "0x1"
      >       },
      > @@ -973,4 +961,4 @@
      >           "SampleAfterValue": "100003",
      >           "UMask": "0x10"
      >       }
      > -]
      > \ No newline at end of file
      > +]
      > diff --git
     a/tools/perf/pmu-events/arch/x86/broadwellx/floating-point.json
     b/tools/perf/pmu-events/arch/x86/broadwellx/floating-point.json
      > index 9ad37dddb354..93bbc8600321 100644
      > --- a/tools/perf/pmu-events/arch/x86/broadwellx/floating-point.json
      > +++ b/tools/perf/pmu-events/arch/x86/broadwellx/floating-point.json
      > @@ -5,6 +5,7 @@
      >           "CounterHTOff": "0,1,2,3",
      >           "EventCode": "0xc7",
      >           "EventName": "FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE",
      > +        "PublicDescription": "Number of SSE/AVX computational
     128-bit packed double precision floating-point instructions retired;
     some instructions will count twice as noted below.  Each count
     represents 2 computation operations, one for each element.  Applies
     to SSE* and AVX* packed double precision floating-point
     instructions: ADD SUB HADD HSUB SUBADD MUL DIV MIN MAX SQRT DPP
     FM(N)ADD/SUB.  DPP and FM(N)ADD/SUB instructions count twice as they
     perform 2 calculations per element. The DAZ and FTZ flags in the
     MXCSR register need to be set when using these events.",
      >           "SampleAfterValue": "2000003",
      >           "UMask": "0x4"
      >       },
      > @@ -14,6 +15,7 @@
      >           "CounterHTOff": "0,1,2,3",
      >           "EventCode": "0xc7",
      >           "EventName": "FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE",
      > +        "PublicDescription": "Number of SSE/AVX computational
     128-bit packed single precision floating-point instructions retired;
     some instructions will count twice as noted below.  Each count
     represents 4 computation operations, one for each element.  Applies
     to SSE* and AVX* packed single precision floating-point
     instructions: ADD SUB HADD HSUB SUBADD MUL DIV MIN MAX SQRT RSQRT
     RCP DPP FM(N)ADD/SUB.  DPP and FM(N)ADD/SUB instructions count twice
     as they perform 2 calculations per element. The DAZ and FTZ flags in
     the MXCSR register need to be set when using these events.",
      >           "SampleAfterValue": "2000003",
      >           "UMask": "0x8"
      >       },
      > @@ -23,6 +25,7 @@
      >           "CounterHTOff": "0,1,2,3",
      >           "EventCode": "0xc7",
      >           "EventName": "FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE",
      > +        "PublicDescription": "Number of SSE/AVX computational
     256-bit packed double precision floating-point instructions retired;
     some instructions will count twice as noted below.  Each count
     represents 4 computation operations, one for each element.  Applies
     to SSE* and AVX* packed double precision floating-point
     instructions: ADD SUB HADD HSUB SUBADD MUL DIV MIN MAX SQRT
     FM(N)ADD/SUB.  FM(N)ADD/SUB instructions count twice as they perform
     2 calculations per element. The DAZ and FTZ flags in the MXCSR
     register need to be set when using these events.",
      >           "SampleAfterValue": "2000003",
      >           "UMask": "0x10"
      >       },
      > @@ -32,6 +35,7 @@
      >           "CounterHTOff": "0,1,2,3",
      >           "EventCode": "0xc7",
      >           "EventName": "FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE",
      > +        "PublicDescription": "Number of SSE/AVX computational
     256-bit packed single precision floating-point instructions retired;
     some instructions will count twice as noted below.  Each count
     represents 8 computation operations, one for each element.  Applies
     to SSE* and AVX* packed single precision floating-point
     instructions: ADD SUB HADD HSUB SUBADD MUL DIV MIN MAX SQRT RSQRT
     RCP DPP FM(N)ADD/SUB.  DPP and FM(N)ADD/SUB instructions count twice
     as they perform 2 calculations per element. The DAZ and FTZ flags in
     the MXCSR register need to be set when using these events.",
      >           "SampleAfterValue": "2000003",
      >           "UMask": "0x20"
      >       },
      > @@ -59,6 +63,7 @@
      >           "CounterHTOff": "0,1,2,3",
      >           "EventCode": "0xc7",
      >           "EventName": "FP_ARITH_INST_RETIRED.SCALAR",
      > +        "PublicDescription": "Number of SSE/AVX computational
     scalar single precision and double precision floating-point
     instructions retired; some instructions will count twice as noted
     below.  Each count represents 1 computational operation. Applies to
     SSE* and AVX* scalar single precision floating-point instructions:
     ADD SUB MUL DIV MIN MAX SQRT RSQRT RCP FM(N)ADD/SUB.  FM(N)ADD/SUB
     instructions count twice as they perform 2 calculations per element.
     The DAZ and FTZ flags in the MXCSR register need to be set when
     using these events.",
      >           "SampleAfterValue": "2000003",
      >           "UMask": "0x3"
      >       },
      > @@ -68,6 +73,7 @@
      >           "CounterHTOff": "0,1,2,3",
      >           "EventCode": "0xc7",
      >           "EventName": "FP_ARITH_INST_RETIRED.SCALAR_DOUBLE",
      > +        "PublicDescription": "Number of SSE/AVX computational
     scalar double precision floating-point instructions retired; some
     instructions will count twice as noted below.  Each count represents
     1 computational operation. Applies to SSE* and AVX* scalar double
     precision floating-point instructions: ADD SUB MUL DIV MIN MAX SQRT
     FM(N)ADD/SUB.  FM(N)ADD/SUB instructions count twice as they perform
     2 calculations per element. The DAZ and FTZ flags in the MXCSR
     register need to be set when using these events.",
      >           "SampleAfterValue": "2000003",
      >           "UMask": "0x1"
      >       },
      > @@ -77,6 +83,7 @@
      >           "CounterHTOff": "0,1,2,3",
      >           "EventCode": "0xc7",
      >           "EventName": "FP_ARITH_INST_RETIRED.SCALAR_SINGLE",
      > +        "PublicDescription": "Number of SSE/AVX computational
     scalar single precision floating-point instructions retired; some
     instructions will count twice as noted below.  Each count represents
     1 computational operation. Applies to SSE* and AVX* scalar single
     precision floating-point instructions: ADD SUB MUL DIV MIN MAX SQRT
     RSQRT RCP FM(N)ADD/SUB.  FM(N)ADD/SUB instructions count twice as
     they perform 2 calculations per element. The DAZ and FTZ flags in
     the MXCSR register need to be set when using these events.",
      >           "SampleAfterValue": "2000003",
      >           "UMask": "0x2"
      >       },
      > @@ -190,4 +197,4 @@
      >           "SampleAfterValue": "2000003",
      >           "UMask": "0x3"
      >       }
      > -]
      > \ No newline at end of file
      > +]
      > diff --git
     a/tools/perf/pmu-events/arch/x86/broadwellx/frontend.json
     b/tools/perf/pmu-events/arch/x86/broadwellx/frontend.json
      > index f0bcb945ff76..37ce8034b2ed 100644
      > --- a/tools/perf/pmu-events/arch/x86/broadwellx/frontend.json
      > +++ b/tools/perf/pmu-events/arch/x86/broadwellx/frontend.json
      > @@ -292,4 +292,4 @@
      >           "SampleAfterValue": "2000003",
      >           "UMask": "0x1"
      >       }
      > -]
      > \ No newline at end of file
      > +]
      > diff --git
     a/tools/perf/pmu-events/arch/x86/broadwellx/memory.json
     b/tools/perf/pmu-events/arch/x86/broadwellx/memory.json
      > index cce993b197e3..545f61f691b9 100644
      > --- a/tools/perf/pmu-events/arch/x86/broadwellx/memory.json
      > +++ b/tools/perf/pmu-events/arch/x86/broadwellx/memory.json
      > @@ -247,7 +247,6 @@
      >           "MSRIndex": "0x1a6,0x1a7",
      >           "MSRValue": "0x3FBFC00244",
      >           "Offcore": "1",
      > -        "PublicDescription": "Counts all demand & prefetch code
     reads miss in the L3",
      >           "SampleAfterValue": "100003",
      >           "UMask": "0x1"
      >       },
      > @@ -258,9 +257,8 @@
      >           "EventCode": "0xB7, 0xBB",
      >           "EventName":
     "OFFCORE_RESPONSE.ALL_CODE_RD.LLC_MISS.LOCAL_DRAM",
      >           "MSRIndex": "0x1a6,0x1a7",
      > -        "MSRValue": "0x0604000244",
      > +        "MSRValue": "0x604000244",
      >           "Offcore": "1",
      > -        "PublicDescription": "Counts all demand & prefetch code
     reads miss the L3 and the data is returned from local dram",
      >           "SampleAfterValue": "100003",
      >           "UMask": "0x1"
      >       },
      > @@ -273,7 +271,6 @@
      >           "MSRIndex": "0x1a6,0x1a7",
      >           "MSRValue": "0x3FBFC00091",
      >           "Offcore": "1",
      > -        "PublicDescription": "Counts all demand & prefetch data
     reads miss in the L3",
      >           "SampleAfterValue": "100003",
      >           "UMask": "0x1"
      >       },
      > @@ -284,9 +281,8 @@
      >           "EventCode": "0xB7, 0xBB",
      >           "EventName":
     "OFFCORE_RESPONSE.ALL_DATA_RD.LLC_MISS.LOCAL_DRAM",
      >           "MSRIndex": "0x1a6,0x1a7",
      > -        "MSRValue": "0x0604000091",
      > +        "MSRValue": "0x604000091",
      >           "Offcore": "1",
      > -        "PublicDescription": "Counts all demand & prefetch data
     reads miss the L3 and the data is returned from local dram",
      >           "SampleAfterValue": "100003",
      >           "UMask": "0x1"
      >       },
      > @@ -297,9 +293,8 @@
      >           "EventCode": "0xB7, 0xBB",
      >           "EventName":
     "OFFCORE_RESPONSE.ALL_DATA_RD.LLC_MISS.REMOTE_DRAM",
      >           "MSRIndex": "0x1a6,0x1a7",
      > -        "MSRValue": "0x063BC00091",
      > +        "MSRValue": "0x63BC00091",
      >           "Offcore": "1",
      > -        "PublicDescription": "Counts all demand & prefetch data
     reads miss the L3 and the data is returned from remote dram",
      >           "SampleAfterValue": "100003",
      >           "UMask": "0x1"
      >       },
      > @@ -312,7 +307,6 @@
      >           "MSRIndex": "0x1a6,0x1a7",
      >           "MSRValue": "0x103FC00091",
      >           "Offcore": "1",
      > -        "PublicDescription": "Counts all demand & prefetch data
     reads miss the L3 and the modified data is transferred from remote
     cache",
      >           "SampleAfterValue": "100003",
      >           "UMask": "0x1"
      >       },
      > @@ -323,9 +317,8 @@
      >           "EventCode": "0xB7, 0xBB",
      >           "EventName":
     "OFFCORE_RESPONSE.ALL_DATA_RD.LLC_MISS.REMOTE_HIT_FORWARD",
      >           "MSRIndex": "0x1a6,0x1a7",
      > -        "MSRValue": "0x087FC00091",
      > +        "MSRValue": "0x87FC00091",
      >           "Offcore": "1",
      > -        "PublicDescription": "Counts all demand & prefetch data
     reads miss the L3 and clean or shared data is transferred from
     remote cache",
      >           "SampleAfterValue": "100003",
      >           "UMask": "0x1"
      >       },
      > @@ -338,20 +331,18 @@
      >           "MSRIndex": "0x1a6,0x1a7",
      >           "MSRValue": "0x3FBFC007F7",
      >           "Offcore": "1",
      > -        "PublicDescription": "Counts all data/code/rfo reads
     (demand & prefetch) miss in the L3",
      >           "SampleAfterValue": "100003",
      >           "UMask": "0x1"
      >       },
      >       {
      > -        "BriefDescription": "Counts all data/code/rfo reads
     (demand & prefetch)miss the L3 and the data is returned from local
     dram",
      > +        "BriefDescription": "Counts all data/code/rfo reads
     (demand & prefetch) miss the L3 and the data is returned from local
     dram",
      >           "Counter": "0,1,2,3",
      >           "CounterHTOff": "0,1,2,3",
      >           "EventCode": "0xB7, 0xBB",
      >           "EventName":
     "OFFCORE_RESPONSE.ALL_READS.LLC_MISS.LOCAL_DRAM",
      >           "MSRIndex": "0x1a6,0x1a7",
      > -        "MSRValue": "0x06040007F7",
      > +        "MSRValue": "0x6040007F7",
      >           "Offcore": "1",
      > -        "PublicDescription": "Counts all data/code/rfo reads
     (demand & prefetch)miss the L3 and the data is returned from local
     dram",
      >           "SampleAfterValue": "100003",
      >           "UMask": "0x1"
      >       },
      > @@ -362,9 +353,8 @@
      >           "EventCode": "0xB7, 0xBB",
      >           "EventName":
     "OFFCORE_RESPONSE.ALL_READS.LLC_MISS.REMOTE_DRAM",
      >           "MSRIndex": "0x1a6,0x1a7",
      > -        "MSRValue": "0x063BC007F7",
      > +        "MSRValue": "0x63BC007F7",
      >           "Offcore": "1",
      > -        "PublicDescription": "Counts all data/code/rfo reads
     (demand & prefetch) miss the L3 and the data is returned from remote
     dram",
      >           "SampleAfterValue": "100003",
      >           "UMask": "0x1"
      >       },
      > @@ -377,7 +367,6 @@
      >           "MSRIndex": "0x1a6,0x1a7",
      >           "MSRValue": "0x103FC007F7",
      >           "Offcore": "1",
      > -        "PublicDescription": "Counts all data/code/rfo reads
     (demand & prefetch) miss the L3 and the modified data is transferred
     from remote cache",
      >           "SampleAfterValue": "100003",
      >           "UMask": "0x1"
      >       },
      > @@ -388,9 +377,8 @@
      >           "EventCode": "0xB7, 0xBB",
      >           "EventName":
     "OFFCORE_RESPONSE.ALL_READS.LLC_MISS.REMOTE_HIT_FORWARD",
      >           "MSRIndex": "0x1a6,0x1a7",
      > -        "MSRValue": "0x087FC007F7",
      > +        "MSRValue": "0x87FC007F7",
      >           "Offcore": "1",
      > -        "PublicDescription": "Counts all data/code/rfo reads
     (demand & prefetch) miss the L3 and clean or shared data is
     transferred from remote cache",
      >           "SampleAfterValue": "100003",
      >           "UMask": "0x1"
      >       },
      > @@ -403,7 +391,6 @@
      >           "MSRIndex": "0x1a6,0x1a7",
      >           "MSRValue": "0x3FBFC08FFF",
      >           "Offcore": "1",
      > -        "PublicDescription": "Counts all requests miss in the L3",
      >           "SampleAfterValue": "100003",
      >           "UMask": "0x1"
      >       },
      > @@ -416,7 +403,6 @@
      >           "MSRIndex": "0x1a6,0x1a7",
      >           "MSRValue": "0x3FBFC00122",
      >           "Offcore": "1",
      > -        "PublicDescription": "Counts all demand & prefetch RFOs
     miss in the L3",
      >           "SampleAfterValue": "100003",
      >           "UMask": "0x1"
      >       },
      > @@ -427,9 +413,8 @@
Quoted 9 lines
      >           "EventCode": "0xB7, 0xBB",
      >           "EventName":
     "OFFCORE_RESPONSE.ALL_RFO.LLC_MISS.LOCAL_DRAM",
      >           "MSRIndex": "0x1a6,0x1a7",
      > -        "MSRValue": "0x0604000122",
      > +        "MSRValue": "0x604000122",
      >           "Offcore": "1",
      > -        "PublicDescription": "Counts all demand & pref

--
Zhengjun Xing

Signature · 1 line
Zhengjun Xing
Signature · 1 line
- Arnaldo