[PATCH v3 0/5] Make evlist CPUs more accurate

From: Ian Rogers <irogers@google.com>
To: Peter Zijlstra <peterz@infradead.org>,
	Ingo Molnar <mingo@redhat.com>,
	Arnaldo Carvalho de Melo <acme@kernel.org>,
	Mark Rutland <mark.rutland@arm.com>,
	Alexander Shishkin <alexander.shishkin@linux.intel.com>,
	Jiri Olsa <jolsa@kernel.org>, Namhyung Kim <namhyung@kernel.org>,
	Mathieu Poirier <mathieu.poirier@linaro.org>,
	Suzuki K Poulose <suzuki.poulose@arm.com>,
	Mike Leach <mike.leach@linaro.org>, Leo Yan <leo.yan@linaro.org>,
	John Garry <john.garry@huawei.com>, Will Deacon <will@kernel.org>,
	Alexei Starovoitov <ast@kernel.org>,
	Daniel Borkmann <daniel@iogearbox.net>,
	Andrii Nakryiko <andrii@kernel.org>,
	Martin KaFai Lau <kafai@fb.com>, Song Liu <songliubraving@fb.com>,
	Yonghong Song <yhs@fb.com>,
	John Fastabend <john.fastabend@gmail.com>,
	KP Singh <kpsingh@kernel.org>, Kajol Jain <kjain@linux.ibm.com>,
	James Clark <james.clark@arm.com>,
	German Gomez <german.gomez@arm.com>,
	Adrian Hunter <adrian.hunter@intel.com>,
	Riccardo Mancini <rickyman7@gmail.com>,
	Andi Kleen <ak@linux.intel.com>,
	Alexey Bayduraev <alexey.v.bayduraev@linux.intel.com>,
	Alexander Antonov <alexander.antonov@linux.intel.com>,
	linux-perf-users@vger.kernel.org, linux-kernel@vger.kernel.org,
	coresight@lists.linaro.org, linux-arm-kernel@lists.infradead.org,
	netdev@vger.kernel.org, bpf@vger.kernel.org
Cc: Stephane Eranian <eranian@google.com>, Ian Rogers <irogers@google.com>
Subject: [PATCH v3 0/5] Make evlist CPUs more accurate
Date: Thu,  7 Apr 2022 20:56:11 -0700	[thread overview]
Message-ID: <20220408035616.1356953-1-irogers@google.com> (raw)

evlist has all_cpus, computed to be the merge of all evsel CPU maps,
and cpus. cpus may contain more CPUs than all_cpus, as by default cpus
holds all online CPUs whilst all_cpus holds the merge/union from
evsels. For an uncore event there may just be 1 CPU per socket, which
will be a far smaller CPU map than all online CPUs.

The v1 patches changed cpus to be called user_requested_cpus, to
reflect their potential user specified nature. The user_requested_cpus
are set to be the current value intersected with all_cpus, so that
user_requested_cpus is always a subset of all_cpus. This fixes
printing code for metrics so that unnecessary blank lines aren't
printed.

To make the intersect function perform well, a perf_cpu_map__is_subset
function is added. While adding this function, the v2 patches also
used it in perf_cpu_map__merge to avoid creating a new CPU map for
some currently missed patterns. The reference counts for these
functions is simplified as discussed here:
https://lore.kernel.org/lkml/YkdOpJDnknrOPq2t@kernel.org/ but this
means users of perf_cpu_map__merge must now do a put on the 1st
argument.

v2. Reorders the "Avoid segv" patch and makes other adjustments
    suggested by Arnaldo Carvalho de Melo <acme@kernel.org>.
v3. Modify reference count behaviour for merge and intersect. Add
    intersect tests and tidy thee cpu map tests suite.

Ian Rogers (5):
  perf cpumap: Don't decrement refcnt on args to merge
  perf tests: Additional cpumap merge tests
  perf cpumap: Add intersect function.
  perf evlist: Respect all_cpus when setting user_requested_cpus
  perf test: Combine cpu map tests into 1 suite

 tools/lib/perf/cpumap.c              | 46 ++++++++++++++---
 tools/lib/perf/evlist.c              |  6 ++-
 tools/lib/perf/include/perf/cpumap.h |  2 +
 tools/perf/tests/builtin-test.c      |  4 +-
 tools/perf/tests/cpumap.c            | 74 +++++++++++++++++++++++++---
 tools/perf/tests/tests.h             |  4 +-
 tools/perf/util/evlist.c             |  7 +++
 7 files changed, 120 insertions(+), 23 deletions(-)

-- 
2.35.1.1178.g4f1659d476-goog