[NOT-FOR-MERGE-PATCH 3/3] Debug: Alternate old/new version of sched_avg_update calls for profiling

[Frederic Weisbecker]

This way we can measure with the function profiler.

	trace_stat/function0:  sched_avg_update_old                   764    223.936 us      0.293 us        0.017 us
	trace_stat/function0:  sched_avg_update                       764    199.818 us      0.261 us        0.006 us
	trace_stat/function1:  sched_avg_update_old                   252    99.225 us       0.393 us        0.106 us
	trace_stat/function1:  sched_avg_update                       252    59.080 us       0.234 us        0.008 us
	trace_stat/function2:  sched_avg_update_old                   606    202.465 us      0.334 us        0.007 us
	trace_stat/function2:  sched_avg_update                       605    184.083 us      0.304 us        0.002 us
	trace_stat/function3:  sched_avg_update_old                   902    261.790 us      0.290 us        0.010 us
	trace_stat/function3:  sched_avg_update                       901    238.253 us      0.264 us        0.004 us
	trace_stat/function4:  sched_avg_update_old                   120    50.391 us       0.419 us        0.095 us
	trace_stat/function4:  sched_avg_update                       120    35.947 us       0.299 us        0.004 us
	trace_stat/function5:  sched_avg_update_old                   581    195.280 us      0.336 us        0.011 us
	trace_stat/function5:  sched_avg_update                       582    173.594 us      0.298 us        0.003 us
	trace_stat/function6:  sched_avg_update_old                   202    45.201 us       0.223 us        0.049 us
	trace_stat/function6:  sched_avg_update                       201    37.892 us       0.188 us        0.013 us
	trace_stat/function7:  sched_avg_update_old                   200    80.731 us       0.403 us        0.477 us
	trace_stat/function7:  sched_avg_update                       200    41.759 us       0.208 us        0.012 us

Cc: Mike Galbraith <efault@xxxxxx>
Cc: Peter Zijlstra <peterz@xxxxxxxxxxxxx>
Cc: Thomas Gleixner <tglx@xxxxxxxxxxxxx>
Cc: Ingo Molnar <mingo@xxxxxxxxxx>
Not-Signed-off-by: Frederic Weisbecker <fweisbec@xxxxxxxxx>
---
 kernel/sched/core.c  | 16 ++++++++++++++++
 kernel/sched/fair.c  |  7 ++++++-
 kernel/sched/sched.h |  8 +++++++-
 3 files changed, 29 insertions(+), 2 deletions(-)

diff --git a/kernel/sched/core.c b/kernel/sched/core.c
index 0c0578a..5410e1bb 100644
--- a/kernel/sched/core.c
+++ b/kernel/sched/core.c
@@ -663,6 +663,22 @@ bool sched_can_stop_tick(struct rq *rq)
 }
 #endif /* CONFIG_NO_HZ_FULL */
 
+void sched_avg_update_old(struct rq *rq)
+{
+	s64 period = sched_avg_period();
+
+	while ((s64)(rq_clock(rq) - rq->age_stamp) > period) {
+		/*
+		 * Inline assembly required to prevent the compiler
+		 * optimising this loop into a divmod call.
+		 * See __iter_div_u64_rem() for another example of this.
+		 */
+		asm("" : "+rm" (rq->age_stamp));
+		rq->age_stamp += period;
+		rq->rt_avg /= 2;
+	}
+}
+
 void sched_avg_update(struct rq *rq)
 {
 	s64 period = sched_avg_period();
diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
index c6dd8ba..1b487bb 100644
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -4607,6 +4607,7 @@ decay_load_missed(unsigned long load, unsigned long missed_updates, int idx)
  * see decay_load_misses(). For NOHZ_FULL we get to subtract and add the extra
  * term.
  */
+DEFINE_PER_CPU(int, cpu_lodd);
 static void cpu_load_update(struct rq *this_rq, unsigned long this_load,
 			    unsigned long pending_updates)
 {
@@ -4647,7 +4648,11 @@ static void cpu_load_update(struct rq *this_rq, unsigned long this_load,
 		this_rq->cpu_load[i] = (old_load * (scale - 1) + new_load) >> i;
 	}
 
-	sched_avg_update(this_rq);
+	if (__this_cpu_read(cpu_lodd) % 2)
+		sched_avg_update(this_rq);
+	else
+		sched_avg_update_old(this_rq);
+	__this_cpu_add(cpu_lodd, 1);
 }
 
 /* Used instead of source_load when we know the type == 0 */
diff --git a/kernel/sched/sched.h b/kernel/sched/sched.h
index 72f1f30..b7e197b 100644
--- a/kernel/sched/sched.h
+++ b/kernel/sched/sched.h
@@ -1427,6 +1427,7 @@ static inline int hrtick_enabled(struct rq *rq)
 #endif /* CONFIG_SCHED_HRTICK */
 
 #ifdef CONFIG_SMP
+extern void sched_avg_update_old(struct rq *rq);
 extern void sched_avg_update(struct rq *rq);
 
 #ifndef arch_scale_freq_capacity
@@ -1448,10 +1449,15 @@ unsigned long arch_scale_cpu_capacity(struct sched_domain *sd, int cpu)
 }
 #endif
 
+DECLARE_PER_CPU(int, cpu_lodd);
 static inline void sched_rt_avg_update(struct rq *rq, u64 rt_delta)
 {
 	rq->rt_avg += rt_delta * arch_scale_freq_capacity(NULL, cpu_of(rq));
-	sched_avg_update(rq);
+	if (__this_cpu_read(cpu_lodd) % 2)
+		sched_avg_update(rq);
+	else
+		sched_avg_update_old(rq);
+	__this_cpu_add(cpu_lodd, 1);
 }
 #else
 static inline void sched_rt_avg_update(struct rq *rq, u64 rt_delta) { }
-- 
2.7.0