On Thu, May 2, 2024 at 11:15 AM Liang, Kan <kan.liang@linux.intel.com> wrote:
>
>
>
> On 2024-05-02 12:01 a.m., Ian Rogers wrote:
> > By assuming sysfs events are either upper or lower case, the case
> > insensitive event parsing can probe for the existence of files rather
> > then loading all events in a directory. When the event is a json event
> > like inst_retired.any on Intel, this reduces the number of openat
> > calls on a Tigerlake laptop from 325 down to 255.
> >
> > v1 sent as an RFC:
> > https://lore.kernel.org/lkml/20240413040812.4042051-1-irogers@google.com/
> >
> > v2: addresses review feedback from Kan Liang, by updating
> > documentation and adding tests.
> >
> > v3: incorporate feedback from Thomas Richter <tmricht@linux.ibm.com>
> > that s390 event names are all upper case. Do a lower case probe
> > then an upper case probe, make documentation and tests also agree.
> >
> > v4: add checks to write (kernel test robot) and fix a typo.
> >
> > Ian Rogers (6):
> > perf test pmu-events: Make it clearer that pmu-events tests json
> > events
> > perf Document: Sysfs event names must be lower or upper case
> > perf test pmu: Refactor format test and exposed test APIs
> > perf test pmu: Add an eagerly loaded event test
> > perf test pmu: Test all sysfs PMU event names are the same case
> > perf pmu: Assume sysfs events are always the same case
> >
> > .../sysfs-bus-event_source-devices-events | 6 +
> > tools/perf/tests/pmu-events.c | 2 +-
> > tools/perf/tests/pmu.c | 468 ++++++++++++------
> > tools/perf/util/parse-events.c | 2 +-
> > tools/perf/util/parse-events.h | 2 +-
> > tools/perf/util/pmu.c | 111 +++--
> > tools/perf/util/pmu.h | 4 +-
> > tools/perf/util/pmus.c | 16 +-
> > tools/perf/util/pmus.h | 2 +
> > 9 files changed, 416 insertions(+), 197 deletions(-)
> >
>
> Except the warning for the uninitialized 'ret', the rest looks good to me.
Thanks Kan, I'll hunt that down to fix in v5 and add your reviewed-by.
Would be awesome to get Thomas or someone else's s390 thumbs up. The
differences between v3 to v5 should be very minor, so I'd accept a
check on any of them :-)
Ian
> Reviewed-by: Kan Liang <kan.liang@linux.intel.com>
>
> Thanks,
> Kan