From: Ingo Molnar <mingo@elte.hu>
This very simple patch temporarily turns off the array-switching within the
current scheduler.
This yields an unfair scheduler (higher-prio CPU loops can starve lower-prio
CPU loops indefinitely), but it gives us most of the benefits of a single
array and gives us a chance to test the effects in isolation.
It would be nice if people could test: basic interactivity and various
SMP/NUMA workloads such as SDET.
Signed-off-by: Andrew Morton <akpm@osdl.org>
---
25-akpm/kernel/sched.c | 6 +++++-
1 files changed, 5 insertions(+), 1 deletion(-)
diff -puN kernel/sched.c~sched-single-array kernel/sched.c
--- 25/kernel/sched.c~sched-single-array 2004-08-06 02:53:01.127320168 -0700
+++ 25-akpm/kernel/sched.c 2004-08-06 03:12:16.015750440 -0700
@@ -2448,7 +2448,11 @@ void scheduler_tick(int user_ticks, int
if (!rq->expired_timestamp)
rq->expired_timestamp = jiffies;
- if (!TASK_INTERACTIVE(p) || EXPIRED_STARVING(rq)) {
+ /*
+ * Disable array switching temporarily to investigate
+ * the effects (note that this doesnt guarantee fairness):
+ */
+ if (0 && (!TASK_INTERACTIVE(p) || EXPIRED_STARVING(rq))) {
enqueue_task(p, rq->expired);
if (p->static_prio < rq->best_expired_prio)
rq->best_expired_prio = p->static_prio;
_