Re: [PATCH v7 0/6] perf stat affinity changes

[Ian Rogers]

On Fri, Feb 6, 2026 at 1:35 PM Arnaldo Carvalho de Melo <acme@xxxxxxxxxx> wrote:
>
> On Tue, Feb 03, 2026 at 02:51:23PM -0800, Ian Rogers wrote:
> > Change how affinities work with evlist__for_each_cpu. Move the
> > affinity code into the iterator to simplify setting it up. Detect when
> > affinities will and won't be profitable, for example a tool event and
> > a regular perf event (or read group) may face less delay from a single
> > IPI for the event read than from a call to sched_setaffinity. Add a
> >  --no-affinity flag to perf stat to allow affinities to be disabled.
> >
> > v7: Revert "perf tool_pmu: More accurately set the cpus for tool
> >     events" that caused issues with user specified CPUs (Andres Freund
> >     <andres@xxxxxxxxxxx>). Fix a null test is prepare_metric so that
> >     missing events can't trigger segfaults (Andres Freund). Make the
> >     CPU map propagation improve the CPU maps for tool events that only
> >     read on index 0, this allows later setting when
> >     evlist__create_maps is called with the correct user CPUs. Rebase
> >     previous non-merged affinity changes that hadn't been picked up
> >     yet.
>
> Not applying to tmp.perf-tools-next, can you please check?

Will send a rebase on tmp.perf-tools-next shortly. Could I interest
you in some second servings of build improvements while you wait? :-)
https://lore.kernel.org/lkml/20260203164323.3447826-1-irogers@xxxxxxxxxx/

Thanks,
Ian

> - Arnaldo
>
> > v6: Drop merged tool event change. Move TPEBS fix into its own patch
> >     1st.
> > https://lore.kernel.org/lkml/20260108212652.768875-1-irogers@xxxxxxxxxx/
> >
> > v5: Drop merged changes. Move tool event reading to first
> >     patch. Change --no-affinity flag to --affinity/--no-affinity flag.
> > https://lore.kernel.org/lkml/20251118211326.1840989-1-irogers@xxxxxxxxxx/
> >     On v5 there was discussion with Andi Kleen who points out that
> >     affinities will work better with real time priorities but using
> >     this requires privileges.
> >
> > v4: Rebase. Add patch to reduce scope of walltime_nsec_stats now that
> >     the legacy metric code is no more. Minor tweak to the ru_stats
> >     clean up.
> > https://lore.kernel.org/lkml/20251113180517.44096-1-irogers@xxxxxxxxxx/
> >
> > v3: Add affinity clean ups and read tool events last.
> > https://lore.kernel.org/lkml/20251106071241.141234-1-irogers@xxxxxxxxxx/
> >
> > v2: Fixed an aggregation index issue:
> > https://lore.kernel.org/lkml/20251104234148.3103176-2-irogers@xxxxxxxxxx/
> >
> > v1:
> > https://lore.kernel.org/lkml/20251104053449.1208800-1-irogers@xxxxxxxxxx/
> >
> > Ian Rogers (6):
> >   Revert "perf tool_pmu: More accurately set the cpus for tool events"
> >   perf stat-shadow: In prepare_metric fix guard on reading NULL
> >     perf_stat_evsel
> >   perf evlist: Special map propagation for tool events that read on 1
> >     CPU
> >   perf evlist: Missing TPEBS close in evlist__close
> >   perf evlist: Reduce affinity use and move into iterator, fix no
> >     affinity
> >   perf stat: Add no-affinity flag
> >
> >  tools/lib/perf/evlist.c                 |  36 +++++-
> >  tools/lib/perf/include/internal/evsel.h |   2 +
> >  tools/perf/Documentation/perf-stat.txt  |   4 +
> >  tools/perf/builtin-stat.c               | 114 ++++++++---------
> >  tools/perf/util/evlist.c                | 156 +++++++++++++++---------
> >  tools/perf/util/evlist.h                |  27 ++--
> >  tools/perf/util/parse-events.c          |  10 +-
> >  tools/perf/util/pmu.c                   |  23 ++++
> >  tools/perf/util/pmu.h                   |   3 +
> >  tools/perf/util/stat-shadow.c           |   7 +-
> >  tools/perf/util/tool_pmu.c              |  19 ---
> >  tools/perf/util/tool_pmu.h              |   1 -
> >  12 files changed, 237 insertions(+), 165 deletions(-)
> >
> > --
> > 2.53.0.rc2.204.g2597b5adb4-goog