Hello,
Currently the interactive points a process can have are in a [-5, 5] range,
that is, 25% of the [0, 39] range. Two reasons are mentionned:
1) nice +19 interactive tasks do not preempt nice 0 CPU hogs.
2) nice -20 CPU hogs do not get preempted by nice 0 tasks.
But, using 50% of the range, instead of 25% the interactivity points are better
spread and both rules are still respected. Having a larger range for
interactivity points it's easier to choose between two interactive tasks.
So, why not changing PRIO_BONUS_RATIO to 50 instead of 25?
Actually it should be in the [45, 49] range to maximize the bonus points
range and satisfy both rules due to integer arithmetic.
Something like that:
--- linux-2.5.74-mm2-O3/kernel/sched.c 2003-07-07 18:46:29.000000000 +0200
+++ linux-2.5.74-mm2-O3/kernel/sched.c-bonus 2003-07-08 15:27:12.000000000 +0200
@@ -71,7 +71,7 @@
#define CHILD_PENALTY 80
#define PARENT_PENALTY 100
#define EXIT_WEIGHT 3
-#define PRIO_BONUS_RATIO 25
+#define PRIO_BONUS_RATIO 45
#define INTERACTIVE_DELTA 2
#define MIN_SLEEP_AVG (HZ)
#define MAX_SLEEP_AVG (10*HZ)
@@ -90,13 +90,13 @@
* We scale it linearly, offset by the INTERACTIVE_DELTA delta.
* Here are a few examples of different nice levels:
*
- * TASK_INTERACTIVE(-20): [1,1,1,1,1,1,1,1,1,0,0]
- * TASK_INTERACTIVE(-10): [1,1,1,1,1,1,1,0,0,0,0]
- * TASK_INTERACTIVE( 0): [1,1,1,1,0,0,0,0,0,0,0]
- * TASK_INTERACTIVE( 10): [1,1,0,0,0,0,0,0,0,0,0]
- * TASK_INTERACTIVE( 19): [0,0,0,0,0,0,0,0,0,0,0]
+ * TASK_INTERACTIVE(-20): [1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,0,0]
+ * TASK_INTERACTIVE(-10): [1,1,1,1,1,1,1,1,1,1,1,1,0,0,0,0,0,0,0]
+ * TASK_INTERACTIVE( 0): [1,1,1,1,1,1,1,1,0,0,0,0,0,0,0,0,0,0,0]
+ * TASK_INTERACTIVE( 10): [1,1,1,1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0]
+ * TASK_INTERACTIVE( 19): [0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0]
*
- * (the X axis represents the possible -5 ... 0 ... +5 dynamic
+ * (the X axis represents the possible -9 ... 0 ... +9 dynamic
* priority range a task can explore, a value of '1' means the
* task is rated interactive.)
*
@@ -325,9 +325,9 @@
* priority but is modified by bonuses/penalties.
*
* We scale the actual sleep average [0 .... MAX_SLEEP_AVG]
- * into the -5 ... 0 ... +5 bonus/penalty range.
+ * into the -9 ... 0 ... +9 bonus/penalty range.
*
- * We use 25% of the full 0...39 priority range so that:
+ * We use 50% of the full 0...39 priority range so that:
*
* 1) nice +19 interactive tasks do not preempt nice 0 CPU hogs.
* 2) nice -20 CPU hogs do not get preempted by nice 0 tasks.
And if you want to try other values for PRIO_BONUS_RATIO, I attached a simple
hack to generate the infos in the above comment.
Another thing that I was wondering is: should every absence on the runqueue be
considered interactive bonus? For example, TASK_UNINTERRIBLE tasks receive
bonus when they wake up. This implies that when a CPU hog becomes a memory hog
and starts swapping, it is considered interactive. OTOH when a task is swapping
I would like it to consume its data the earliest possible, to avoid losing the
swapping benefit.
So I'd like to know if the patch below is a good or bad thing.
--- linux-2.5.74-mm2-O3/kernel/sched.c 2003-07-07 18:46:29.000000000 +0200
+++ linux-2.5.74-mm2-O3/kernel/sched.c-INTERR 2003-07-08 17:43:59.000000000 +0200
@@ -388,7 +388,7 @@ static inline void activate_task(task_t
{
long sleep_time = jiffies - p->last_run - 1;
- if (sleep_time > 0) {
+ if (sleep_time > 0 && p->state == TASK_INTERRUPTIBLE) {
unsigned long runtime = jiffies - p->avg_start;
/*
Thanks for your wisdom.
Guillaume
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Tue Jul 15 2003 - 22:00:28 EST