sched: revise boost logic when boost_type is SCHED_BOOST_ON_BIG

At present HMP scheduler boost tends to pack tasks by taking into
account of power cost and cstate.  This is suboptimal to performance
as it can lead preemption and higher latency.

Revise logic to prefer the least loaded CPU among the big cluster CPUs
when boost type is SCHED_BOOST_ON_BIG.  New logic still honor the
behaviour that scheduler can place tasks on the little CPUs when the
big CPUs are all overcommitted.

Also, it's found that need_idle with boost can easily return previous
CPU when there is no idle CPU found.  Fix this issue by making
need_idle flag to take precedence over sched_boost.

CRs-fixed: 1074879
Change-Id: I470bcd0588e038b4a540d337fe6a412f2fa74920
Signed-off-by: Joonwoo Park <joonwoop@codeaurora.org>
Signed-off-by: Syed Rameez Mustafa <rameezmustafa@codeaurora.org>

1 files changed, 33 insertions, 3 deletions

diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
index 1674b1054f83..40c2e7ac69e0 100644
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -2596,6 +2596,7 @@ static u32 __compute_runnable_contrib(u64 n)
 #define SBC_FLAG_COLOC_CLUSTER				0x10000
 #define SBC_FLAG_WAKER_CLUSTER				0x20000
 #define SBC_FLAG_BACKUP_CLUSTER				0x40000
+#define SBC_FLAG_BOOST_CLUSTER				0x80000
 
 struct cpu_select_env {
 	struct task_struct *p;
@@ -2708,6 +2709,18 @@ select_least_power_cluster(struct cpu_select_env *env)
 		env->task_load = scale_load_to_cpu(task_load(env->p),
 			cluster_first_cpu(env->rtg->preferred_cluster));
 		env->sbc_best_cluster_flag |= SBC_FLAG_COLOC_CLUSTER;
+
+		if (env->boost_type != SCHED_BOOST_NONE) {
+			for_each_sched_cluster(cluster) {
+				if (cluster != env->rtg->preferred_cluster) {
+					__set_bit(cluster->id,
+						env->backup_list);
+					__clear_bit(cluster->id,
+						env->candidate_list);
+				}
+			}
+		}
+
 		return env->rtg->preferred_cluster;
 	}
 
@@ -2961,7 +2974,13 @@ static void find_best_cpu_in_cluster(struct sched_cluster *c,
 		update_spare_capacity(stats, env, i, c->capacity,
 				      env->cpu_load);
 
-		if (env->boost_type == SCHED_BOOST_ON_ALL ||
+		/*
+		 * need_idle takes precedence over sched boost but when both
+		 * are set, idlest CPU with in all the clusters is selected
+		 * when boost_type = BOOST_ON_ALL whereas idlest CPU in the
+		 * big cluster is selected within boost_type = BOOST_ON_BIG.
+		 */
+		if ((!env->need_idle && env->boost_type != SCHED_BOOST_NONE) ||
 		    env->need_waker_cluster ||
 		    sched_cpu_high_irqload(i) ||
 		    spill_threshold_crossed(env, cpu_rq(i)))
@@ -3178,12 +3197,23 @@ retry:
 		sbc_flag |= env.sbc_best_flag;
 		target = stats.best_cpu;
 	} else {
-		if (env.rtg) {
+		if (env.rtg && env.boost_type == SCHED_BOOST_NONE) {
 			env.rtg = NULL;
 			goto retry;
 		}
 
-		find_backup_cluster(&env, &stats);
+		/*
+		 * With boost_type == SCHED_BOOST_ON_BIG, we reach here with
+		 * backup_list = little cluster, candidate_list = none and
+		 * stats->best_capacity_cpu points the best spare capacity
+		 * CPU among the CPUs in the big cluster.
+		 */
+		if (env.boost_type == SCHED_BOOST_ON_BIG &&
+		    stats.best_capacity_cpu >= 0)
+			sbc_flag |= SBC_FLAG_BOOST_CLUSTER;
+		else
+			find_backup_cluster(&env, &stats);
+
 		if (stats.best_capacity_cpu >= 0) {
 			target = stats.best_capacity_cpu;
 			sbc_flag |= SBC_FLAG_BEST_CAP_CPU;