Commit 6668128a authored by Peter Zijlstra's avatar Peter Zijlstra Committed by Ingo Molnar

perf/core: Optimize ctx_sched_out()

When an event group contains more events than can be scheduled on the
hardware, iterating the full event group for ctx_sched_out is a waste
of time.

Keep track of the events that got programmed on the hardware, such
that we can iterate this smaller list in order to schedule them out.
Signed-off-by: default avatarPeter Zijlstra (Intel) <peterz@infradead.org>
Acked-by: default avatarMark Rutland <mark.rutland@arm.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Alexey Budankov <alexey.budankov@linux.intel.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: David Carrillo-Cisneros <davidcc@google.com>
Cc: Dmitri Prokhorov <Dmitry.Prohorov@intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Kan Liang <kan.liang@intel.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: Valery Cherepennikov <valery.cherepennikov@intel.com>
Cc: Vince Weaver <vincent.weaver@maine.edu>
Cc: linux-kernel@vger.kernel.org
Signed-off-by: default avatarIngo Molnar <mingo@kernel.org>
parent 8343aae6
......@@ -553,6 +553,7 @@ struct perf_event {
* either sufficies for read.
*/
struct list_head sibling_list;
struct list_head active_list;
/*
* Node on the pinned or flexible tree located at the event context;
*/
......@@ -718,6 +719,10 @@ struct perf_event_context {
struct perf_event_groups pinned_groups;
struct perf_event_groups flexible_groups;
struct list_head event_list;
struct list_head pinned_active;
struct list_head flexible_active;
int nr_events;
int nr_active;
int is_active;
......
......@@ -1647,14 +1647,6 @@ perf_event_groups_rotate(struct perf_event_groups *groups, int cpu)
typeof(*event), node); event; \
event = rb_entry_safe(rb_next(&event->node), \
typeof(*event), node))
/*
* Iterate event groups with cpu == key.
*/
#define perf_event_groups_for_each_cpu(event, key, groups, node) \
for (event = perf_event_groups_first(groups, key); \
event && event->cpu == key; \
event = rb_entry_safe(rb_next(&event->node), \
typeof(*event), node))
/*
* Add a event from the lists for its context.
......@@ -1889,8 +1881,9 @@ list_del_event(struct perf_event *event, struct perf_event_context *ctx)
static void perf_group_detach(struct perf_event *event)
{
struct perf_event *sibling, *tmp;
struct perf_event_context *ctx = event->ctx;
lockdep_assert_held(&event->ctx->lock);
lockdep_assert_held(&ctx->lock);
/*
* We can have double detach due to exit/hot-unplug + close.
......@@ -1924,6 +1917,13 @@ static void perf_group_detach(struct perf_event *event)
if (!RB_EMPTY_NODE(&event->group_node)) {
list_del_init(&sibling->sibling_list);
add_event_to_groups(sibling, event->ctx);
if (sibling->state == PERF_EVENT_STATE_ACTIVE) {
struct list_head *list = sibling->attr.pinned ?
&ctx->pinned_active : &ctx->flexible_active;
list_add_tail(&sibling->active_list, list);
}
}
WARN_ON_ONCE(sibling->ctx != event->ctx);
......@@ -1988,6 +1988,13 @@ event_sched_out(struct perf_event *event,
if (event->state != PERF_EVENT_STATE_ACTIVE)
return;
/*
* Asymmetry; we only schedule events _IN_ through ctx_sched_in(), but
* we can schedule events _OUT_ individually through things like
* __perf_remove_from_context().
*/
list_del_init(&event->active_list);
perf_pmu_disable(event->pmu);
event->pmu->del(event, 0);
......@@ -2835,9 +2842,8 @@ static void ctx_sched_out(struct perf_event_context *ctx,
struct perf_cpu_context *cpuctx,
enum event_type_t event_type)
{
int sw = -1, cpu = smp_processor_id();
struct perf_event *event, *tmp;
int is_active = ctx->is_active;
struct perf_event *event;
lockdep_assert_held(&ctx->lock);
......@@ -2884,20 +2890,12 @@ static void ctx_sched_out(struct perf_event_context *ctx,
perf_pmu_disable(ctx->pmu);
if (is_active & EVENT_PINNED) {
perf_event_groups_for_each_cpu(event, cpu,
&ctx->pinned_groups, group_node)
group_sched_out(event, cpuctx, ctx);
perf_event_groups_for_each_cpu(event, sw,
&ctx->pinned_groups, group_node)
list_for_each_entry_safe(event, tmp, &ctx->pinned_active, active_list)
group_sched_out(event, cpuctx, ctx);
}
if (is_active & EVENT_FLEXIBLE) {
perf_event_groups_for_each_cpu(event, cpu,
&ctx->flexible_groups, group_node)
group_sched_out(event, cpuctx, ctx);
perf_event_groups_for_each_cpu(event, sw,
&ctx->flexible_groups, group_node)
list_for_each_entry_safe(event, tmp, &ctx->flexible_active, active_list)
group_sched_out(event, cpuctx, ctx);
}
perf_pmu_enable(ctx->pmu);
......@@ -3231,8 +3229,10 @@ static int pinned_sched_in(struct perf_event *event, void *data)
if (!event_filter_match(event))
return 0;
if (group_can_go_on(event, sid->cpuctx, sid->can_add_hw))
group_sched_in(event, sid->cpuctx, sid->ctx);
if (group_can_go_on(event, sid->cpuctx, sid->can_add_hw)) {
if (!group_sched_in(event, sid->cpuctx, sid->ctx))
list_add_tail(&event->active_list, &sid->ctx->pinned_active);
}
/*
* If this pinned group hasn't been scheduled,
......@@ -3255,7 +3255,9 @@ static int flexible_sched_in(struct perf_event *event, void *data)
return 0;
if (group_can_go_on(event, sid->cpuctx, sid->can_add_hw)) {
if (group_sched_in(event, sid->cpuctx, sid->ctx))
if (!group_sched_in(event, sid->cpuctx, sid->ctx))
list_add_tail(&event->active_list, &sid->ctx->flexible_active);
else
sid->can_add_hw = 0;
}
......@@ -3973,6 +3975,8 @@ static void __perf_event_init_context(struct perf_event_context *ctx)
perf_event_groups_init(&ctx->pinned_groups);
perf_event_groups_init(&ctx->flexible_groups);
INIT_LIST_HEAD(&ctx->event_list);
INIT_LIST_HEAD(&ctx->pinned_active);
INIT_LIST_HEAD(&ctx->flexible_active);
atomic_set(&ctx->refcount, 1);
}
......@@ -9815,6 +9819,7 @@ perf_event_alloc(struct perf_event_attr *attr, int cpu,
INIT_LIST_HEAD(&event->event_entry);
INIT_LIST_HEAD(&event->sibling_list);
INIT_LIST_HEAD(&event->active_list);
init_event_group(event);
INIT_LIST_HEAD(&event->rb_entry);
INIT_LIST_HEAD(&event->active_entry);
......
Markdown is supported
0%
or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment