Hi Ian,
On Tue, Nov 14, 2023 at 1:13 PM Ian Rogers <[email protected]> wrote:
>
> When the cycles event isn't available evsel will fallback to the
> cpu-clock software event. task-clock is similar to cpu-clock but only
> runs when the process is running. Falling back to cpu-clock when not
> system wide leads to confusion, by falling back to task-clock it is
> hoped the confusion is less.
I think they are almost the same and no meaningful difference.
The cpu-clock event also runs only when the task is running if
it's a per-task event.
>
> Update a nearby comment and debug string for the change.
>
> Signed-off-by: Ian Rogers <[email protected]>
> ---
> tools/perf/util/evsel.c | 15 ++++++++-------
> 1 file changed, 8 insertions(+), 7 deletions(-)
>
> diff --git a/tools/perf/util/evsel.c b/tools/perf/util/evsel.c
> index a5da74e3a517..e1175313e4bc 100644
> --- a/tools/perf/util/evsel.c
> +++ b/tools/perf/util/evsel.c
> @@ -2861,18 +2861,19 @@ bool evsel__fallback(struct evsel *evsel, int err, char *msg, size_t msgsize)
> evsel->core.attr.type == PERF_TYPE_HARDWARE &&
> evsel->core.attr.config == PERF_COUNT_HW_CPU_CYCLES) {
> /*
> - * If it's cycles then fall back to hrtimer based
> - * cpu-clock-tick sw counter, which is always available even if
> - * no PMU support.
> + * If it's cycles then fall back to hrtimer based cpu-clock sw
> + * counter, which is always available even if no PMU support.
> *
> * PPC returns ENXIO until 2.6.37 (behavior changed with commit
> * b0a873e).
> */
> - scnprintf(msg, msgsize, "%s",
> -"The cycles event is not supported, trying to fall back to cpu-clock-ticks");
> -
> evsel->core.attr.type = PERF_TYPE_SOFTWARE;
> - evsel->core.attr.config = PERF_COUNT_SW_CPU_CLOCK;
> + evsel->core.attr.config = evsel->core.system_wide
I'm not sure you can use the system_wide flag for this.
IIUC it's to override the target setting in some cases
(e.g. a dummy event to track sideband events in all CPUs)
and I think you need to check target__has_cpu() instead.
But as I said above, it won't make any difference in the
output. Conceptually it'd be more natural to use task-clock
event for per-task sessions though.
Thanks,
Namhyung
> + ? PERF_COUNT_SW_CPU_CLOCK
> + : PERF_COUNT_SW_TASK_CLOCK;
> + scnprintf(msg, msgsize,
> + "The cycles event is not supported, trying to fall back to %s",
> + evsel->core.system_wide ? "cpu-clock" : "task-clock");
>
> zfree(&evsel->name);
> return true;
> --
> 2.43.0.rc0.421.g78406f8d94-goog
>