Skip to contentSkip to conversation

sched/numa: Introduce alternative wakeup bias

tip:sched/numa tip-bot for Peter Zijlstra 1 participant 1 message Oct 12, 2012 Scheduler

lore
Share & tools
lore
#0 lore raw
Commit-ID:  52a30f1fa3b9f0c0d1270b4e34be865e83b3e31b
Gitweb:     http://git.kernel.org/tip/52a30f1fa3b9f0c0d1270b4e34be865e83b3e31b
Author:     Peter Zijlstra <[email protected]>
AuthorDate: Fri, 5 Oct 2012 15:57:54 +0200
Committer:  Ingo Molnar <[email protected]>
CommitDate: Fri, 12 Oct 2012 12:07:16 +0200

sched/numa: Introduce alternative wakeup bias

Rename NUMA_BIAS to NUMA_TTWU_BIAS to clarify what it does.

Also, disable by default, it seems too agressive. Also provide an alternative to play with, instead of altering the prev cpu, alter the waking cpu, maybe that's less agressive.

No clear data either way for the moment.

---
 kernel/sched/fair.c     |    7 +++++--
 kernel/sched/features.h |    3 ++-
 2 files changed, 7 insertions(+), 3 deletions(-)
kernel/sched/fair.c +5 -2
index f586f17..62d77ef 100644
@@ -3009,7 +3009,7 @@ select_task_rq_fair(struct task_struct *p, int sd_flag, int wake_flags)
 	}
 
 	rcu_read_lock();
-	if (sched_feat_numa(NUMA_BIAS) && node != -1) {
+	if (sched_feat_numa(NUMA_TTWU_BIAS) && node != -1) {
 		/*
 		 * For fork,exec find the idlest cpu in the home-node.
 		 */
@@ -3031,7 +3031,10 @@ select_task_rq_fair(struct task_struct *p, int sd_flag, int wake_flags)
 			if (node_cpu < 0)
 				goto find_sd;
 
-			prev_cpu = node_cpu;
+			if (sched_feat_numa(NUMA_TTWU_TO))
+				cpu = node_cpu;
+			else
+				prev_cpu = node_cpu;
 		}
 	}
 
kernel/sched/features.h +2 -1
index 845b838..3293af4 100644
@@ -66,7 +66,8 @@ SCHED_FEAT(LB_MIN, false)
 SCHED_FEAT(NUMA,           true)
 SCHED_FEAT(NUMA_FORCE_BIG, false)
 SCHED_FEAT(NUMA_HOT,       true)
-SCHED_FEAT(NUMA_BIAS,      true)
+SCHED_FEAT(NUMA_TTWU_BIAS, false)
+SCHED_FEAT(NUMA_TTWU_TO,   false)
 SCHED_FEAT(NUMA_PULL,      true)
 SCHED_FEAT(NUMA_PULL_BIAS, true)
 #endif