sched: Define structure to store the sched_group statistics for fbg()
authorGautham R Shenoy <ego@in.ibm.com>
Wed, 25 Mar 2009 09:13:46 +0000 (14:43 +0530)
committerIngo Molnar <mingo@elte.hu>
Wed, 25 Mar 2009 09:30:45 +0000 (10:30 +0100)
Impact: cleanup

Currently a whole bunch of variables are used to store the
various statistics pertaining to the groups we iterate over
in find_busiest_group().

Group them together in a single data structure and add
appropriate comments.

This will be useful later on when we create helper functions
to calculate the sched_group statistics.

Credit: Vaidyanathan Srinivasan <svaidy@linux.vnet.ibm.com>
Signed-off-by: Gautham R Shenoy <ego@in.ibm.com>
Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Suresh Siddha <suresh.b.siddha@intel.com>
Cc: "Balbir Singh" <balbir@in.ibm.com>
Cc: Nick Piggin <nickpiggin@yahoo.com.au>
Cc: "Dhaval Giani" <dhaval@linux.vnet.ibm.com>
Cc: Bharata B Rao <bharata@linux.vnet.ibm.com>
LKML-Reference: <20090325091345.13992.20099.stgit@sofia.in.ibm.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
kernel/sched.c

index f87adbe999e06f09f8b81d46f8d8ac567adc9bd9..109db122de50d614f53145a21060773f7dbc772b 100644 (file)
@@ -3191,6 +3191,18 @@ static int move_one_task(struct rq *this_rq, int this_cpu, struct rq *busiest,
 }
 /********** Helpers for find_busiest_group ************************/
 
+/**
+ * sg_lb_stats - stats of a sched_group required for load_balancing
+ */
+struct sg_lb_stats {
+       unsigned long avg_load; /*Avg load across the CPUs of the group */
+       unsigned long group_load; /* Total load over the CPUs of the group */
+       unsigned long sum_nr_running; /* Nr tasks running in the group */
+       unsigned long sum_weighted_load; /* Weighted load of group's tasks */
+       unsigned long group_capacity;
+       int group_imb; /* Is there an imbalance in the group ? */
+};
+
 /**
  * group_first_cpu - Returns the first cpu in the cpumask of a sched_group.
  * @group: The group whose first cpu is to be returned.
@@ -3257,23 +3269,22 @@ find_busiest_group(struct sched_domain *sd, int this_cpu,
        load_idx = get_sd_load_idx(sd, idle);
 
        do {
-               unsigned long load, group_capacity, max_cpu_load, min_cpu_load;
+               struct sg_lb_stats sgs;
+               unsigned long load, max_cpu_load, min_cpu_load;
                int local_group;
                int i;
-               int __group_imb = 0;
                unsigned int balance_cpu = -1, first_idle_cpu = 0;
-               unsigned long sum_nr_running, sum_weighted_load;
                unsigned long sum_avg_load_per_task;
                unsigned long avg_load_per_task;
 
                local_group = cpumask_test_cpu(this_cpu,
                                               sched_group_cpus(group));
+               memset(&sgs, 0, sizeof(sgs));
 
                if (local_group)
                        balance_cpu = group_first_cpu(group);
 
                /* Tally up the load of all CPUs in the group */
-               sum_weighted_load = sum_nr_running = avg_load = 0;
                sum_avg_load_per_task = avg_load_per_task = 0;
 
                max_cpu_load = 0;
@@ -3301,9 +3312,9 @@ find_busiest_group(struct sched_domain *sd, int this_cpu,
                                        min_cpu_load = load;
                        }
 
-                       avg_load += load;
-                       sum_nr_running += rq->nr_running;
-                       sum_weighted_load += weighted_cpuload(i);
+                       sgs.group_load += load;
+                       sgs.sum_nr_running += rq->nr_running;
+                       sgs.sum_weighted_load += weighted_cpuload(i);
 
                        sum_avg_load_per_task += cpu_avg_load_per_task(i);
                }
@@ -3320,12 +3331,12 @@ find_busiest_group(struct sched_domain *sd, int this_cpu,
                        goto ret;
                }
 
-               total_load += avg_load;
+               total_load += sgs.group_load;
                total_pwr += group->__cpu_power;
 
                /* Adjust by relative CPU power of the group */
-               avg_load = sg_div_cpu_power(group,
-                               avg_load * SCHED_LOAD_SCALE);
+               sgs.avg_load = sg_div_cpu_power(group,
+                               sgs.group_load * SCHED_LOAD_SCALE);
 
 
                /*
@@ -3341,22 +3352,23 @@ find_busiest_group(struct sched_domain *sd, int this_cpu,
                                sum_avg_load_per_task * SCHED_LOAD_SCALE);
 
                if ((max_cpu_load - min_cpu_load) > 2*avg_load_per_task)
-                       __group_imb = 1;
+                       sgs.group_imb = 1;
 
-               group_capacity = group->__cpu_power / SCHED_LOAD_SCALE;
+               sgs.group_capacity = group->__cpu_power / SCHED_LOAD_SCALE;
 
                if (local_group) {
-                       this_load = avg_load;
+                       this_load = sgs.avg_load;
                        this = group;
-                       this_nr_running = sum_nr_running;
-                       this_load_per_task = sum_weighted_load;
-               } else if (avg_load > max_load &&
-                          (sum_nr_running > group_capacity || __group_imb)) {
-                       max_load = avg_load;
+                       this_nr_running = sgs.sum_nr_running;
+                       this_load_per_task = sgs.sum_weighted_load;
+               } else if (sgs.avg_load > max_load &&
+                          (sgs.sum_nr_running > sgs.group_capacity ||
+                               sgs.group_imb)) {
+                       max_load = sgs.avg_load;
                        busiest = group;
-                       busiest_nr_running = sum_nr_running;
-                       busiest_load_per_task = sum_weighted_load;
-                       group_imb = __group_imb;
+                       busiest_nr_running = sgs.sum_nr_running;
+                       busiest_load_per_task = sgs.sum_weighted_load;
+                       group_imb = sgs.group_imb;
                }
 
 #if defined(CONFIG_SCHED_MC) || defined(CONFIG_SCHED_SMT)
@@ -3372,7 +3384,7 @@ find_busiest_group(struct sched_domain *sd, int this_cpu,
                 * If the local group is idle or completely loaded
                 * no need to do power savings balance at this domain
                 */
-               if (local_group && (this_nr_running >= group_capacity ||
+               if (local_group && (this_nr_running >= sgs.group_capacity ||
                                    !this_nr_running))
                        power_savings_balance = 0;
 
@@ -3380,8 +3392,9 @@ find_busiest_group(struct sched_domain *sd, int this_cpu,
                 * If a group is already running at full capacity or idle,
                 * don't include that group in power savings calculations
                 */
-               if (!power_savings_balance || sum_nr_running >= group_capacity
-                   || !sum_nr_running)
+               if (!power_savings_balance ||
+                       sgs.sum_nr_running >= sgs.group_capacity ||
+                       !sgs.sum_nr_running)
                        goto group_next;
 
                /*
@@ -3389,13 +3402,13 @@ find_busiest_group(struct sched_domain *sd, int this_cpu,
                 * This is the group from where we need to pick up the load
                 * for saving power
                 */
-               if ((sum_nr_running < min_nr_running) ||
-                   (sum_nr_running == min_nr_running &&
+               if ((sgs.sum_nr_running < min_nr_running) ||
+                   (sgs.sum_nr_running == min_nr_running &&
                     group_first_cpu(group) > group_first_cpu(group_min))) {
                        group_min = group;
-                       min_nr_running = sum_nr_running;
-                       min_load_per_task = sum_weighted_load /
-                                               sum_nr_running;
+                       min_nr_running = sgs.sum_nr_running;
+                       min_load_per_task = sgs.sum_weighted_load /
+                                               sgs.sum_nr_running;
                }
 
                /*
@@ -3403,14 +3416,14 @@ find_busiest_group(struct sched_domain *sd, int this_cpu,
                 * capacity but still has some space to pick up some load
                 * from other group and save more power
                 */
-               if (sum_nr_running > group_capacity - 1)
+               if (sgs.sum_nr_running > sgs.group_capacity - 1)
                        goto group_next;
 
-               if (sum_nr_running > leader_nr_running ||
-                   (sum_nr_running == leader_nr_running &&
+               if (sgs.sum_nr_running > leader_nr_running ||
+                   (sgs.sum_nr_running == leader_nr_running &&
                     group_first_cpu(group) < group_first_cpu(group_leader))) {
                        group_leader = group;
-                       leader_nr_running = sum_nr_running;
+                       leader_nr_running = sgs.sum_nr_running;
                }
 group_next:
 #endif