hollalinux
/
linux-sg2042
mirror of https://github.com/sophgo/linux-riscv.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

sched: Define structure to store the sched_domain statistics for fbg() 

Impact: cleanup

Currently we use a lot of local variables in find_busiest_group()
to capture the various statistics related to the sched_domain.
Group them together into a single data structure.

This will help us to offload the job of updating the sched_domain
statistics to a helper function.

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: <20090325091356.13992.25970.stgit@sofia.in.ibm.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
This commit is contained in:
Gautham R Shenoy 2009-03-25 14:43:56 +05:30 committed by Ingo Molnar
parent 1f8c553d0f
commit 222d656dea
1 changed files with 120 additions and 85 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

205
kernel/sched.c
View File

@ -3190,6 +3190,37 @@ static int move_one_task(struct rq *this_rq, int this_cpu, struct rq *busiest,
return 0; return 0;
} }
/********** Helpers for find_busiest_group ************************/ /********** Helpers for find_busiest_group ************************/
/**
* sd_lb_stats - Structure to store the statistics of a sched_domain
* during load balancing.
*/
struct sd_lb_stats {
struct sched_group *busiest; /* Busiest group in this sd */
struct sched_group *this; /* Local group in this sd */
unsigned long total_load; /* Total load of all groups in sd */
unsigned long total_pwr; /* Total power of all groups in sd */
unsigned long avg_load; /* Average load across all groups in sd */
/** Statistics of this group */
unsigned long this_load;
unsigned long this_load_per_task;
unsigned long this_nr_running;
/* Statistics of the busiest group */
unsigned long max_load;
unsigned long busiest_load_per_task;
unsigned long busiest_nr_running;
int group_imb; /* Is there imbalance in this sd */
#if defined(CONFIG_SCHED_MC) || defined(CONFIG_SCHED_SMT)
int power_savings_balance; /* Is powersave balance needed for this sd */
struct sched_group *group_min; /* Least loaded group in sd */
struct sched_group *group_leader; /* Group which relieves group_min */
unsigned long min_load_per_task; /* load_per_task in group_min */
unsigned long leader_nr_running; /* Nr running of group_leader */
unsigned long min_nr_running; /* Nr running of group_min */
#endif
};
/** /**
* sg_lb_stats - stats of a sched_group required for load_balancing * sg_lb_stats - stats of a sched_group required for load_balancing
@ -3346,23 +3377,16 @@ find_busiest_group(struct sched_domain *sd, int this_cpu,
unsigned long *imbalance, enum cpu_idle_type idle, unsigned long *imbalance, enum cpu_idle_type idle,
int *sd_idle, const struct cpumask *cpus, int *balance) int *sd_idle, const struct cpumask *cpus, int *balance)
{ {
struct sched_group *busiest = NULL, *this = NULL, *group = sd->groups; struct sd_lb_stats sds;
unsigned long max_load, avg_load, total_load, this_load, total_pwr; struct sched_group *group = sd->groups;
unsigned long max_pull; unsigned long max_pull;
unsigned long busiest_load_per_task, busiest_nr_running; int load_idx;
unsigned long this_load_per_task, this_nr_running;
int load_idx, group_imb = 0; memset(&sds, 0, sizeof(sds));
#if defined(CONFIG_SCHED_MC) || defined(CONFIG_SCHED_SMT) #if defined(CONFIG_SCHED_MC) || defined(CONFIG_SCHED_SMT)
int power_savings_balance = 1; sds.power_savings_balance = 1;
unsigned long leader_nr_running = 0, min_load_per_task = 0; sds.min_nr_running = ULONG_MAX;
unsigned long min_nr_running = ULONG_MAX;
struct sched_group *group_min = NULL, *group_leader = NULL;
#endif #endif
max_load = this_load = total_load = total_pwr = 0;
busiest_load_per_task = busiest_nr_running = 0;
this_load_per_task = this_nr_running = 0;
load_idx = get_sd_load_idx(sd, idle); load_idx = get_sd_load_idx(sd, idle);
do { do {
@ -3378,22 +3402,22 @@ find_busiest_group(struct sched_domain *sd, int this_cpu,
if (balance && !(*balance)) if (balance && !(*balance))
goto ret; goto ret;
total_load += sgs.group_load; sds.total_load += sgs.group_load;
total_pwr += group->__cpu_power; sds.total_pwr += group->__cpu_power;
if (local_group) { if (local_group) {
this_load = sgs.avg_load; sds.this_load = sgs.avg_load;
this = group; sds.this = group;
this_nr_running = sgs.sum_nr_running; sds.this_nr_running = sgs.sum_nr_running;
this_load_per_task = sgs.sum_weighted_load; sds.this_load_per_task = sgs.sum_weighted_load;
} else if (sgs.avg_load > max_load && } else if (sgs.avg_load > sds.max_load &&
(sgs.sum_nr_running > sgs.group_capacity || (sgs.sum_nr_running > sgs.group_capacity ||
sgs.group_imb)) { sgs.group_imb)) {
max_load = sgs.avg_load; sds.max_load = sgs.avg_load;
busiest = group; sds.busiest = group;
busiest_nr_running = sgs.sum_nr_running; sds.busiest_nr_running = sgs.sum_nr_running;
busiest_load_per_task = sgs.sum_weighted_load; sds.busiest_load_per_task = sgs.sum_weighted_load;
group_imb = sgs.group_imb; sds.group_imb = sgs.group_imb;
} }
#if defined(CONFIG_SCHED_MC) || defined(CONFIG_SCHED_SMT) #if defined(CONFIG_SCHED_MC) || defined(CONFIG_SCHED_SMT)
@ -3409,15 +3433,16 @@ find_busiest_group(struct sched_domain *sd, int this_cpu,
* If the local group is idle or completely loaded * If the local group is idle or completely loaded
* no need to do power savings balance at this domain * no need to do power savings balance at this domain
*/ */
if (local_group && (this_nr_running >= sgs.group_capacity || if (local_group &&
!this_nr_running)) (sds.this_nr_running >= sgs.group_capacity ||
power_savings_balance = 0; !sds.this_nr_running))
sds.power_savings_balance = 0;
/* /*
* If a group is already running at full capacity or idle, * If a group is already running at full capacity or idle,
* don't include that group in power savings calculations * don't include that group in power savings calculations
*/ */
if (!power_savings_balance || if (!sds.power_savings_balance ||
sgs.sum_nr_running >= sgs.group_capacity || sgs.sum_nr_running >= sgs.group_capacity ||
!sgs.sum_nr_running) !sgs.sum_nr_running)
goto group_next; goto group_next;
@ -3427,12 +3452,13 @@ find_busiest_group(struct sched_domain *sd, int this_cpu,
* This is the group from where we need to pick up the load * This is the group from where we need to pick up the load
* for saving power * for saving power
*/ */
if ((sgs.sum_nr_running < min_nr_running) || if ((sgs.sum_nr_running < sds.min_nr_running) ||
(sgs.sum_nr_running == min_nr_running && (sgs.sum_nr_running == sds.min_nr_running &&
group_first_cpu(group) > group_first_cpu(group_min))) { group_first_cpu(group) >
group_min = group; group_first_cpu(sds.group_min))) {
min_nr_running = sgs.sum_nr_running; sds.group_min = group;
min_load_per_task = sgs.sum_weighted_load / sds.min_nr_running = sgs.sum_nr_running;
sds.min_load_per_task = sgs.sum_weighted_load /
sgs.sum_nr_running; sgs.sum_nr_running;
} }
@ -3444,29 +3470,32 @@ find_busiest_group(struct sched_domain *sd, int this_cpu,
if (sgs.sum_nr_running > sgs.group_capacity - 1) if (sgs.sum_nr_running > sgs.group_capacity - 1)
goto group_next; goto group_next;
if (sgs.sum_nr_running > leader_nr_running || if (sgs.sum_nr_running > sds.leader_nr_running ||
(sgs.sum_nr_running == leader_nr_running && (sgs.sum_nr_running == sds.leader_nr_running &&
group_first_cpu(group) < group_first_cpu(group_leader))) { group_first_cpu(group) <
group_leader = group; group_first_cpu(sds.group_leader))) {
leader_nr_running = sgs.sum_nr_running; sds.group_leader = group;
sds.leader_nr_running = sgs.sum_nr_running;
} }
group_next: group_next:
#endif #endif
group = group->next; group = group->next;
} while (group != sd->groups); } while (group != sd->groups);
if (!busiest || this_load >= max_load || busiest_nr_running == 0) if (!sds.busiest || sds.this_load >= sds.max_load
|| sds.busiest_nr_running == 0)
goto out_balanced; goto out_balanced;
avg_load = (SCHED_LOAD_SCALE * total_load) / total_pwr; sds.avg_load = (SCHED_LOAD_SCALE * sds.total_load) / sds.total_pwr;
if (this_load >= avg_load || if (sds.this_load >= sds.avg_load ||
100*max_load <= sd->imbalance_pct*this_load) 100*sds.max_load <= sd->imbalance_pct * sds.this_load)
goto out_balanced; goto out_balanced;
busiest_load_per_task /= busiest_nr_running; sds.busiest_load_per_task /= sds.busiest_nr_running;
if (group_imb) if (sds.group_imb)
busiest_load_per_task = min(busiest_load_per_task, avg_load); sds.busiest_load_per_task =
min(sds.busiest_load_per_task, sds.avg_load);
/* /*
* We're trying to get all the cpus to the average_load, so we don't * We're trying to get all the cpus to the average_load, so we don't
@ -3479,7 +3508,7 @@ group_next:
* by pulling tasks to us. Be careful of negative numbers as they'll * by pulling tasks to us. Be careful of negative numbers as they'll
* appear as very large values with unsigned longs. * appear as very large values with unsigned longs.
*/ */
if (max_load <= busiest_load_per_task) if (sds.max_load <= sds.busiest_load_per_task)
goto out_balanced; goto out_balanced;
/* /*
@ -3487,17 +3516,18 @@ group_next:
* max load less than avg load(as we skip the groups at or below * max load less than avg load(as we skip the groups at or below
* its cpu_power, while calculating max_load..) * its cpu_power, while calculating max_load..)
*/ */
if (max_load < avg_load) { if (sds.max_load < sds.avg_load) {
*imbalance = 0; *imbalance = 0;
goto small_imbalance; goto small_imbalance;
} }
/* Don't want to pull so many tasks that a group would go idle */ /* Don't want to pull so many tasks that a group would go idle */
max_pull = min(max_load - avg_load, max_load - busiest_load_per_task); max_pull = min(sds.max_load - sds.avg_load,
sds.max_load - sds.busiest_load_per_task);
/* How much load to actually move to equalise the imbalance */ /* How much load to actually move to equalise the imbalance */
*imbalance = min(max_pull * busiest->__cpu_power, *imbalance = min(max_pull * sds.busiest->__cpu_power,
(avg_load - this_load) * this->__cpu_power) (sds.avg_load - sds.this_load) * sds.this->__cpu_power)
/ SCHED_LOAD_SCALE; / SCHED_LOAD_SCALE;
/* /*
@ -3506,24 +3536,27 @@ group_next:
* a think about bumping its value to force at least one task to be * a think about bumping its value to force at least one task to be
* moved * moved
*/ */
if (*imbalance < busiest_load_per_task) { if (*imbalance < sds.busiest_load_per_task) {
unsigned long tmp, pwr_now, pwr_move; unsigned long tmp, pwr_now, pwr_move;
unsigned int imbn; unsigned int imbn;
small_imbalance: small_imbalance:
pwr_move = pwr_now = 0; pwr_move = pwr_now = 0;
imbn = 2; imbn = 2;
if (this_nr_running) { if (sds.this_nr_running) {
this_load_per_task /= this_nr_running; sds.this_load_per_task /= sds.this_nr_running;
if (busiest_load_per_task > this_load_per_task) if (sds.busiest_load_per_task >
sds.this_load_per_task)
imbn = 1; imbn = 1;
} else } else
this_load_per_task = cpu_avg_load_per_task(this_cpu); sds.this_load_per_task =
cpu_avg_load_per_task(this_cpu);
if (max_load - this_load + busiest_load_per_task >= if (sds.max_load - sds.this_load +
busiest_load_per_task * imbn) { sds.busiest_load_per_task >=
*imbalance = busiest_load_per_task; sds.busiest_load_per_task * imbn) {
return busiest; *imbalance = sds.busiest_load_per_task;
return sds.busiest;
} }
/* /*
@ -3532,52 +3565,54 @@ small_imbalance:
* moving them. * moving them.
*/ */
pwr_now += busiest->__cpu_power * pwr_now += sds.busiest->__cpu_power *
min(busiest_load_per_task, max_load); min(sds.busiest_load_per_task, sds.max_load);
pwr_now += this->__cpu_power * pwr_now += sds.this->__cpu_power *
min(this_load_per_task, this_load); min(sds.this_load_per_task, sds.this_load);
pwr_now /= SCHED_LOAD_SCALE; pwr_now /= SCHED_LOAD_SCALE;
/* Amount of load we'd subtract */ /* Amount of load we'd subtract */
tmp = sg_div_cpu_power(busiest, tmp = sg_div_cpu_power(sds.busiest,
busiest_load_per_task * SCHED_LOAD_SCALE); sds.busiest_load_per_task * SCHED_LOAD_SCALE);
if (max_load > tmp) if (sds.max_load > tmp)
pwr_move += busiest->__cpu_power * pwr_move += sds.busiest->__cpu_power *
min(busiest_load_per_task, max_load - tmp); min(sds.busiest_load_per_task,
sds.max_load - tmp);
/* Amount of load we'd add */ /* Amount of load we'd add */
if (max_load * busiest->__cpu_power < if (sds.max_load * sds.busiest->__cpu_power <
busiest_load_per_task * SCHED_LOAD_SCALE) sds.busiest_load_per_task * SCHED_LOAD_SCALE)
tmp = sg_div_cpu_power(this, tmp = sg_div_cpu_power(sds.this,
max_load * busiest->__cpu_power); sds.max_load * sds.busiest->__cpu_power);
else else
tmp = sg_div_cpu_power(this, tmp = sg_div_cpu_power(sds.this,
busiest_load_per_task * SCHED_LOAD_SCALE); sds.busiest_load_per_task * SCHED_LOAD_SCALE);
pwr_move += this->__cpu_power * pwr_move += sds.this->__cpu_power *
min(this_load_per_task, this_load + tmp); min(sds.this_load_per_task,
sds.this_load + tmp);
pwr_move /= SCHED_LOAD_SCALE; pwr_move /= SCHED_LOAD_SCALE;
/* Move if we gain throughput */ /* Move if we gain throughput */
if (pwr_move > pwr_now) if (pwr_move > pwr_now)
*imbalance = busiest_load_per_task; *imbalance = sds.busiest_load_per_task;
} }
return busiest; return sds.busiest;
out_balanced: out_balanced:
#if defined(CONFIG_SCHED_MC) || defined(CONFIG_SCHED_SMT) #if defined(CONFIG_SCHED_MC) || defined(CONFIG_SCHED_SMT)
if (idle == CPU_NOT_IDLE || !(sd->flags & SD_POWERSAVINGS_BALANCE)) if (idle == CPU_NOT_IDLE || !(sd->flags & SD_POWERSAVINGS_BALANCE))
goto ret; goto ret;
if (this != group_leader || group_leader == group_min) if (sds.this != sds.group_leader || sds.group_leader == sds.group_min)
goto ret; goto ret;
*imbalance = min_load_per_task; *imbalance = sds.min_load_per_task;
if (sched_mc_power_savings >= POWERSAVINGS_BALANCE_WAKEUP) { if (sched_mc_power_savings >= POWERSAVINGS_BALANCE_WAKEUP) {
cpu_rq(this_cpu)->rd->sched_mc_preferred_wakeup_cpu = cpu_rq(this_cpu)->rd->sched_mc_preferred_wakeup_cpu =
group_first_cpu(group_leader); group_first_cpu(sds.group_leader);
} }
return group_min; return sds.group_min;
#endif #endif
ret: ret: