Commit fe45bafb authored Jan 19, 2017 by Alexander Shishkin Committed by Ingo Molnar Jan 30, 2017

perf/core: Don't re-schedule CPU flexible events needlessly



In the sched-in path, we first remove a CPU's flexible events in order to
give priority to the task's pinned events. However, this step can be safely
skipped if the task doesn't have its own pinned events.

This patch implements this skipping.

Reported-by: Adrian Hunter <adrian.hunter@intel.com>
Signed-off-by: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Arnaldo Carvalho de Melo <acme@infradead.org>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vince Weaver <vincent.weaver@maine.edu>
Cc: vince@deater.net
Link: http://lkml.kernel.org/r/20170119164330.22887-2-alexander.shishkin@linux.intel.com


Signed-off-by: Ingo Molnar <mingo@kernel.org>

parent 1fd7e416

kernel/events/core.c

+5 −1

Original line number	Diff line number	Diff line
		@@ -3122,7 +3122,11 @@ static void perf_event_context_sched_in(struct perf_event_context *ctx,
		* We want to keep the following priority order:
		* cpu pinned (that don't need to move), task pinned,
		* cpu flexible, task flexible.
		*
		* However, if task's ctx is not carrying any pinned
		* events, no need to flip the cpuctx's events around.
		*/
		if (!list_empty(&ctx->pinned_groups))
		cpu_ctx_sched_out(cpuctx, EVENT_FLEXIBLE);
		perf_event_sched_in(cpuctx, ctx, task);
		perf_pmu_enable(ctx->pmu);

Admin message