Patch from Robert Love <rml@tech9.net>

And here goes my current scheduler bits.  Pretty minimal:

	- tuning knob for NUMA's node_threshold
	- better defaults for interactivity:
		o max_timeslice: 300ms -> 200ms
		o default timeslice: 150ms -> 100ms
		o child_penalty: 95 -> 50
		o max_sleep_avg: 2s -> 5s
		o starvation_limit: 2s -> 5s
	- bring back the smart wake up from Ingo's old tree
	- rq_idx() in UP should typecast to long
	- cleanup

Patch is on top of 2.5.59-mm8 with the previous Ingo scheduler patch
added and the old patches removed.



 linux/sysctl.h |    1 +
 sched.c        |   39 ++++++++++++++++++++++++++++-----------
 sysctl.c       |    4 ++++
 3 files changed, 33 insertions(+), 11 deletions(-)

diff -puN include/linux/sysctl.h~rml-scheduler-update2 include/linux/sysctl.h
--- 25/include/linux/sysctl.h~rml-scheduler-update2	2003-02-07 00:58:35.000000000 -0800
+++ 25-akpm/include/linux/sysctl.h	2003-02-07 00:58:35.000000000 -0800
@@ -169,6 +169,7 @@ enum {
 	SCHED_INTERACTIVE_DELTA=7,	/* delta used to scale interactivity */
 	SCHED_MAX_SLEEP_AVG=8,		/* maximum sleep avg attainable */
 	SCHED_STARVATION_LIMIT=9,	/* no re-active if expired is starved */
+	SCHED_NODE_THRESHOLD=10,	/* NUMA node rebalance threshold */
 };
 
 /* CTL_NET names: */
diff -puN kernel/sched.c~rml-scheduler-update2 kernel/sched.c
--- 25/kernel/sched.c~rml-scheduler-update2	2003-02-07 00:58:35.000000000 -0800
+++ 25-akpm/kernel/sched.c	2003-02-07 00:59:12.000000000 -0800
@@ -54,22 +54,23 @@
 /*
  * These are the 'tuning knobs' of the scheduler:
  *
- * Minimum timeslice is 10 msecs, default timeslice is 150 msecs,
- * maximum timeslice is 300 msecs. Timeslices get refilled after
+ * Minimum timeslice is 10 msecs, default timeslice is 100 msecs,
+ * maximum timeslice is 200 msecs. Timeslices get refilled after
  * they expire.
  *
  * They are configurable via /proc/sys/sched
  */
 
 int min_timeslice = (10 * HZ) / 1000;
-int max_timeslice = (300 * HZ) / 1000;
-int child_penalty = 95;
+int max_timeslice = (200 * HZ) / 1000;
+int child_penalty = 50;
 int parent_penalty = 100;
 int exit_weight = 3;
 int prio_bonus_ratio = 25;
 int interactive_delta = 2;
-int max_sleep_avg = 2 * HZ;
-int starvation_limit = 2 * HZ;
+int max_sleep_avg = 5 * HZ;
+int starvation_limit = 3 * HZ;
+int node_threshold = 125;
 
 #define MIN_TIMESLICE		(min_timeslice)
 #define MAX_TIMESLICE		(max_timeslice)
@@ -81,7 +82,9 @@ int starvation_limit = 2 * HZ;
 #define MAX_SLEEP_AVG		(max_sleep_avg)
 #define STARVATION_LIMIT	(starvation_limit)
 #define AGRESSIVE_IDLE_STEAL	1
-#define NODE_THRESHOLD          125
+#define NODE_THRESHOLD          (node_threshold)
+#define SYNC_WAKEUPS		1
+#define SMART_WAKE_CHILD	1
 
 /*
  * If a task is 'interactive' then we reinsert it in the active
@@ -179,7 +182,7 @@ struct prio_array {
 # define cpu_active_balance(c) (cpu_rq(c)->cpu[0].active_balance)
 #else
 # define MAX_NR_SIBLINGS 1
-# define rq_idx(cpu) (cpu)
+# define rq_idx(cpu) ((long) cpu)
 # define cpu_idx(cpu) 0
 # define for_each_sibling(idx, rq) while (0)
 # define cpu_active_balance(c) 0
@@ -573,6 +576,8 @@ static int try_to_wake_up(task_t * p, in
 	long old_state;
 	runqueue_t *rq;
 
+	sync &= SYNC_WAKEUPS;
+
 repeat_lock_task:
 	rq = task_rq_lock(p, &flags);
 	old_state = p->state;
@@ -632,7 +637,19 @@ void wake_up_forked_process(task_t * p)
 		p->prio = effective_prio(p);
 	}
 	set_task_cpu(p, smp_processor_id());
-	activate_task(p, rq);
+
+	if (SMART_WAKE_CHILD) {
+		if (unlikely(!current->array))
+			__activate_task(p, rq);
+		else {
+			p->prio = current->prio;
+			list_add_tail(&p->run_list, &current->run_list);
+			p->array = current->array;
+			p->array->nr_active++;
+			nr_running_inc(rq);
+		}
+	} else
+		activate_task(p, rq);
 
 	rq_unlock(rq);
 }
@@ -2060,7 +2077,7 @@ out_unlock:
  * @user_mask_ptr: user-space pointer to the new cpu mask
  */
 asmlinkage int sys_sched_setaffinity(pid_t pid, unsigned int len,
-				      unsigned long *user_mask_ptr)
+				     unsigned long *user_mask_ptr)
 {
 	unsigned long new_mask;
 	int retval;
@@ -2112,7 +2129,7 @@ out_unlock:
  * @user_mask_ptr: user-space pointer to hold the current cpu mask
  */
 asmlinkage int sys_sched_getaffinity(pid_t pid, unsigned int len,
-				      unsigned long *user_mask_ptr)
+				     unsigned long *user_mask_ptr)
 {
 	unsigned int real_len;
 	unsigned long mask;
diff -puN kernel/sysctl.c~rml-scheduler-update2 kernel/sysctl.c
--- 25/kernel/sysctl.c~rml-scheduler-update2	2003-02-07 00:58:35.000000000 -0800
+++ 25-akpm/kernel/sysctl.c	2003-02-07 00:58:35.000000000 -0800
@@ -64,6 +64,7 @@ extern int prio_bonus_ratio;
 extern int interactive_delta;
 extern int max_sleep_avg;
 extern int starvation_limit;
+extern int node_threshold;
 
 /* this is needed for the proc_dointvec_minmax for [fs_]overflow UID and GID */
 static int maxolduid = 65535;
@@ -399,6 +400,9 @@ static ctl_table sched_table[] = {
 	{SCHED_STARVATION_LIMIT, "starvation_limit", &starvation_limit,
 	 sizeof(int), 0644, NULL, &proc_dointvec_minmax,
 	 &sysctl_intvec, NULL, &zero, NULL},
+	{SCHED_NODE_THRESHOLD, "node_threshold", &node_threshold,
+	 sizeof(int), 0644, NULL, &proc_dointvec_minmax,
+	 &sysctl_intvec, NULL, &one, NULL},
 	{0}
 };
 

_