Hi Arnaldo, Hi Jiri,
Kan's patch series for AlderLake perf core support has been upstreamed, so the interface will not be
changed any more.
For this perf tool series (v4), do you have any comments?
Thanks
Jin Yao
On 4/16/2021 10:04 PM, Jin Yao wrote:
AlderLake uses a hybrid architecture utilizing Golden Cove cores
(core cpu) and Gracemont cores (atom cpu). Each cpu has dedicated
event list. Some events are available on core cpu, some events
are available on atom cpu and some events can be available on both.
Kernel exports new pmus "cpu_core" and "cpu_atom" through sysfs:
/sys/devices/cpu_core
/sys/devices/cpu_atom
cat /sys/devices/cpu_core/cpus
0-15
cat /sys/devices/cpu_atom/cpus
16-23
In this example, core cpus are 0-15 and atom cpus are 16-23.
To enable a core only event or atom only event:
cpu_core/<event name>/
or
cpu_atom/<event name>/
Count the 'cycles' event on core cpus.
# perf stat -e cpu_core/cycles/ -a -- sleep 1
Performance counter stats for 'system wide':
12,853,951,349 cpu_core/cycles/
1.002581249 seconds time elapsed
If one event is available on both atom cpu and core cpu, two events
are created automatically.
# perf stat -e cycles -a -- sleep 1
Performance counter stats for 'system wide':
12,856,467,438 cpu_core/cycles/
6,404,634,785 cpu_atom/cycles/
1.002453013 seconds time elapsed
Group is supported if the events are from same pmu, otherwise a warning
is displayed and disable grouping automatically.
# perf stat -e '{cpu_core/cycles/,cpu_core/instructions/}' -a -- sleep 1
Performance counter stats for 'system wide':
12,863,866,968 cpu_core/cycles/
554,795,017 cpu_core/instructions/
1.002616117 seconds time elapsed
# perf stat -e '{cpu_core/cycles/,cpu_atom/instructions/}' -a -- sleep 1
WARNING: events in group from different hybrid PMUs!
WARNING: grouped events cpus do not match, disabling group:
anon group { cpu_core/cycles/, cpu_atom/instructions/ }
Performance counter stats for 'system wide':
6,283,970 cpu_core/cycles/
765,635 cpu_atom/instructions/
1.003959036 seconds time elapsed
Note that, since the whole patchset for AlderLake hybrid support is very
large (40+ patches). For simplicity, it's splitted into several patch
series.
The patch series 1 only supports the basic functionality. The advanced
supports for perf-c2c/perf-mem/topdown/metrics/topology header and others
will be added in follow-up patch series.
The perf tool codes can also be found at:
https://github.com/yaoj/perf.git
v4:
---
- In Liang Kan's patch:
'[PATCH V6 21/25] perf: Extend PERF_TYPE_HARDWARE and PERF_TYPE_HW_CACHE',
the user interface for hardware events and cache events are changed, so
perf tool patches are changed as well.
- Fix an issue when atom CPUs are offlined.
"/sys/bus/event_source/devices/cpu_atom/cpus"
exists but the content is empty. For this case, we can't enable the cpu_atom
PMU. '[PATCH v4 05/25] perf pmu: Save detected hybrid pmus to a global pmu
list'
- Define 'ret' variable for return value in patch
'[PATCH v4 09/25] perf parse-events: Create two hybrid cache events'
- Directly return add_raw_hybrid() in patch
'[PATCH v4 10/25] perf parse-events: Create two hybrid raw events'
- Drop the patch 'perf pmu: Support 'cycles' and 'branches' inside
hybrid PMU'.
- Separate '[PATCH v3 12/27] perf parse-events: Support no alias assigned event
inside hybrid PMU' into two patches:
'[PATCH v4 11/25] perf parse-events: Compare with hybrid pmu name'
'[PATCH v4 12/25] perf parse-events: Support event inside hybrid pmu'.
And these two patches are improved according to Jiri's comments.
v3:
---
- Drop 'perf evlist: Hybrid event uses its own cpus'. This patch is wide
and actually it's not very necessary. The current perf framework has
processed the cpus for evsel well even for hybrid evsel. So this patch can
be dropped.
- Drop 'perf evsel: Adjust hybrid event and global event mixed group'.
The patch is a bit tricky and hard to understand. In v3, we will disable
grouping when the group members are from different PMUs. So this patch
would be not necessary.
- Create parse-events-hybrid.c/parse-events-hybrid.h and
evlist-hybrid.c/evlist-hybrid.h.
Move hybrid related codes to these files.
- Create a new patch 'perf pmu: Support 'cycles' and 'branches' inside hybrid
PMU' to
support 'cycles' and 'branches' inside PMU.
- Create a new patch 'perf record: Uniquify hybrid event name' to tell user the
pmu which the event belongs to for perf-record.
- If group members are from different hybrid PMUs, shows warning and disable
grouping.
- Other refining and refactoring.
v2:
---
- Drop kernel patches (Kan posted the series "Add Alder Lake support for perf
(kernel)" separately).
- Drop the patches for perf-c2c/perf-mem/topdown/metrics/topology header
supports,
which will be added in series 2 or series 3.
- Simplify the arguments of __perf_pmu__new_alias() by passing
the 'struct pme_event' pointer.
- Check sysfs validity before access.
- Use pmu style event name, such as "cpu_core/cycles/".
- Move command output two chars to the right.
- Move pmu hybrid functions to new created pmu-hybrid.c/pmu-hybrid.h.
This is to pass the perf test python case.
Jin Yao (25):
tools headers uapi: Update tools's copy of linux/perf_event.h
perf jevents: Support unit value "cpu_core" and "cpu_atom"
perf pmu: Simplify arguments of __perf_pmu__new_alias
perf pmu: Save pmu name
perf pmu: Save detected hybrid pmus to a global pmu list
perf pmu: Add hybrid helper functions
perf stat: Uniquify hybrid event name
perf parse-events: Create two hybrid hardware events
perf parse-events: Create two hybrid cache events
perf parse-events: Create two hybrid raw events
perf parse-events: Compare with hybrid pmu name
perf parse-events: Support event inside hybrid pmu
perf record: Create two hybrid 'cycles' events by default
perf stat: Add default hybrid events
perf stat: Filter out unmatched aggregation for hybrid event
perf stat: Warn group events from different hybrid PMU
perf record: Uniquify hybrid event name
perf tests: Add hybrid cases for 'Parse event definition strings' test
perf tests: Add hybrid cases for 'Roundtrip evsel->name' test
perf tests: Skip 'Setup struct perf_event_attr' test for hybrid
perf tests: Support 'Track with sched_switch' test for hybrid
perf tests: Support 'Parse and process metrics' test for hybrid
perf tests: Support 'Session topology' test for hybrid
perf tests: Support 'Convert perf time to TSC' test for hybrid
perf tests: Skip 'perf stat metrics (shadow stat) test' for hybrid
include/uapi/linux/perf_event.h | 15 ++
tools/include/uapi/linux/perf_event.h | 15 ++
tools/perf/builtin-record.c | 47 +++++-
tools/perf/builtin-stat.c | 29 ++++
tools/perf/pmu-events/jevents.c | 2 +
tools/perf/tests/attr.c | 4 +
tools/perf/tests/evsel-roundtrip-name.c | 19 ++-
tools/perf/tests/parse-events.c | 152 ++++++++++++++++++
tools/perf/tests/parse-metric.c | 10 +-
tools/perf/tests/perf-time-to-tsc.c | 16 ++
tools/perf/tests/shell/stat+shadow_stat.sh | 3 +
tools/perf/tests/switch-tracking.c | 10 +-
tools/perf/tests/topology.c | 10 +-
tools/perf/util/Build | 3 +
tools/perf/util/evlist-hybrid.c | 88 ++++++++++
tools/perf/util/evlist-hybrid.h | 14 ++
tools/perf/util/evlist.c | 5 +-
tools/perf/util/evsel.c | 12 +-
tools/perf/util/evsel.h | 4 +-
tools/perf/util/parse-events-hybrid.c | 178 +++++++++++++++++++++
tools/perf/util/parse-events-hybrid.h | 23 +++
tools/perf/util/parse-events.c | 86 +++++++++-
tools/perf/util/parse-events.h | 9 +-
tools/perf/util/parse-events.y | 9 +-
tools/perf/util/pmu-hybrid.c | 89 +++++++++++
tools/perf/util/pmu-hybrid.h | 22 +++
tools/perf/util/pmu.c | 64 +++++---
tools/perf/util/pmu.h | 7 +
tools/perf/util/python-ext-sources | 2 +
tools/perf/util/stat-display.c | 35 +++-
30 files changed, 933 insertions(+), 49 deletions(-)
create mode 100644 tools/perf/util/evlist-hybrid.c
create mode 100644 tools/perf/util/evlist-hybrid.h
create mode 100644 tools/perf/util/parse-events-hybrid.c
create mode 100644 tools/perf/util/parse-events-hybrid.h
create mode 100644 tools/perf/util/pmu-hybrid.c
create mode 100644 tools/perf/util/pmu-hybrid.h
