Based on the thread:
https://lore.kernel.org/linux-perf-users/CAP-5=fVMHzTfKdpWMXtbtx7t14u2f4WzNak+F0Q93cQ7CZfhbg@mail.gmail.com/

First patch is a cherry-pick to avoid a conflict of:
https://lore.kernel.org/linux-perf-users/20220414014642.3308206-2-irogers@google.com/
Second patch makes all_cpus more accurate when there are command line CPUs.
The third patch fixes perf stat metric-only output for uncore metrics.
The fourth patch makes cleans up merging of dummy CPU maps.
The fifth and sixth patch try to make user_requested_cpus and all_cpus
clearer with documentation and by renaming all_cpus.

The code no longer needs to add an intersect function and so the API
is removed and the merged API left unchanged.

v5. Fixes the 1st patch as pointed out by Namhyung Kim
    <namhyung@kernel.org>, adds some further clean-up to it and adds
    reviewed-by Adrian Hunter <adrian.hunter@intel.com> to patch 3.

Ian Rogers (6):
  perf cpumap: Switch to using perf_cpu_map API
  perf evlist: Clear all_cpus before propagating
  perf stat: Avoid printing cpus with no counters
  perf cpumap: Handle dummy maps as empty in subset
  perf evlist: Add to user_requested_cpus documentation
  perf evlist: Rename all_cpus

 tools/lib/perf/cpumap.c                  |  4 +--
 tools/lib/perf/evlist.c                  | 14 +++++---
 tools/lib/perf/include/internal/evlist.h |  5 +--
 tools/perf/builtin-record.c              | 13 +++----
 tools/perf/tests/cpumap.c                | 10 +++++-
 tools/perf/util/bpf_counter_cgroup.c     | 46 +++++++++++-------------
 tools/perf/util/evlist.c                 |  6 ++--
 tools/perf/util/evlist.h                 |  4 +--
 tools/perf/util/stat-display.c           |  7 ++--
 9 files changed, 60 insertions(+), 49 deletions(-)

-- 
2.36.0.464.gb9c8b46e94-goog