AcoSail/kernel_optimize_test
8
0
Fork 0
forked from luck/tmp_suning_uos_patched
Code Pull Requests Activity

perf/core: Optimize perf_rotate_context() event scheduling

Browse Source 
The event schedule order (as per perf_event_sched_in()) is:

 - cpu  pinned
 - task pinned
 - cpu  flexible
 - task flexible

But perf_rotate_context() will unschedule cpu-flexible even if it
doesn't need a rotation.

Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vince Weaver <vincent.weaver@maine.edu>
Cc: linux-kernel@vger.kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
This commit is contained in:
Peter Zijlstra 2018-03-09 14:56:27 +01:00 committed by Ingo Molnar
parent 8703a7cfe1
commit 8d5bce0c37
1 changed files with 37 additions and 23 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

60
kernel/events/core.c
View File

@ -430,7 +430,7 @@ static void update_perf_cpu_limits(void)
WRITE_ONCE(perf_sample_allowed_ns, tmp); WRITE_ONCE(perf_sample_allowed_ns, tmp);
} }
static int perf_rotate_context(struct perf_cpu_context *cpuctx); static bool perf_rotate_context(struct perf_cpu_context *cpuctx);
int perf_proc_update_handler(struct ctl_table *table, int write, int perf_proc_update_handler(struct ctl_table *table, int write,
void __user *buffer, size_t *lenp, void __user *buffer, size_t *lenp,
@ -1041,7 +1041,7 @@ list_update_cgroup_event(struct perf_event *event,
static enum hrtimer_restart perf_mux_hrtimer_handler(struct hrtimer *hr) static enum hrtimer_restart perf_mux_hrtimer_handler(struct hrtimer *hr)
{ {
struct perf_cpu_context *cpuctx; struct perf_cpu_context *cpuctx;
int rotations = 0; bool rotations;
lockdep_assert_irqs_disabled(); lockdep_assert_irqs_disabled();
@ -3600,52 +3600,66 @@ static void rotate_ctx(struct perf_event_context *ctx, struct perf_event *event)
perf_event_groups_insert(&ctx->flexible_groups, event); perf_event_groups_insert(&ctx->flexible_groups, event);
} }
static int perf_rotate_context(struct perf_cpu_context *cpuctx) static inline struct perf_event *
ctx_first_active(struct perf_event_context *ctx)
{ {
struct perf_event *ctx_event = NULL, *cpuctx_event = NULL; return list_first_entry_or_null(&ctx->flexible_active,
struct perf_event, active_list);
}
static bool perf_rotate_context(struct perf_cpu_context *cpuctx)
{
struct perf_event *cpu_event = NULL, *task_event = NULL;
bool cpu_rotate = false, task_rotate = false;
struct perf_event_context *ctx = NULL; struct perf_event_context *ctx = NULL;
int rotate = 0;
/*
* Since we run this from IRQ context, nobody can install new
* events, thus the event count values are stable.
*/
if (cpuctx->ctx.nr_events) { if (cpuctx->ctx.nr_events) {
if (cpuctx->ctx.nr_events != cpuctx->ctx.nr_active) if (cpuctx->ctx.nr_events != cpuctx->ctx.nr_active)
rotate = 1; cpu_rotate = true;
} }
ctx = cpuctx->task_ctx; ctx = cpuctx->task_ctx;
if (ctx && ctx->nr_events) { if (ctx && ctx->nr_events) {
if (ctx->nr_events != ctx->nr_active) if (ctx->nr_events != ctx->nr_active)
rotate = 1; task_rotate = true;
} }
if (!rotate) if (!(cpu_rotate || task_rotate))
goto done; return false;
perf_ctx_lock(cpuctx, cpuctx->task_ctx); perf_ctx_lock(cpuctx, cpuctx->task_ctx);
perf_pmu_disable(cpuctx->ctx.pmu); perf_pmu_disable(cpuctx->ctx.pmu);
cpuctx_event = list_first_entry_or_null(&cpuctx->ctx.flexible_active, if (task_rotate)
struct perf_event, active_list); task_event = ctx_first_active(ctx);
if (ctx) { if (cpu_rotate)
ctx_event = list_first_entry_or_null(&ctx->flexible_active, cpu_event = ctx_first_active(&cpuctx->ctx);
struct perf_event, active_list);
}
cpu_ctx_sched_out(cpuctx, EVENT_FLEXIBLE); /*
if (ctx) * As per the order given at ctx_resched() first 'pop' task flexible
* and then, if needed CPU flexible.
*/
if (task_event || (ctx && cpu_event))
ctx_sched_out(ctx, cpuctx, EVENT_FLEXIBLE); ctx_sched_out(ctx, cpuctx, EVENT_FLEXIBLE);
if (cpu_event)
cpu_ctx_sched_out(cpuctx, EVENT_FLEXIBLE);
if (cpuctx_event) if (task_event)
rotate_ctx(&cpuctx->ctx, cpuctx_event); rotate_ctx(ctx, task_event);
if (ctx_event) if (cpu_event)
rotate_ctx(ctx, ctx_event); rotate_ctx(&cpuctx->ctx, cpu_event);
perf_event_sched_in(cpuctx, ctx, current); perf_event_sched_in(cpuctx, ctx, current);
perf_pmu_enable(cpuctx->ctx.pmu); perf_pmu_enable(cpuctx->ctx.pmu);
perf_ctx_unlock(cpuctx, cpuctx->task_ctx); perf_ctx_unlock(cpuctx, cpuctx->task_ctx);
done:
return rotate; return true;
} }
void perf_event_task_tick(void) void perf_event_task_tick(void)