On 5/27/19 9:13 PM, Peter Zijlstra wrote:
> On Mon, May 27, 2019 at 12:24:07PM -0400, Rik van Riel wrote:
>> On Mon, 2019-05-27 at 07:21 +0100, Dietmar Eggemann wrote:
>>> This is done to align the per cpu (i.e. per rq) load with the util
>>> counterpart (cpu_util(int cpu)). The term 'weighted' is not needed
>>> since there is no 'unweighted' load to distinguish it from.
>>
>> I can see why you want to make cpu_util() and cpu_load()
>> have the same parameter, but ...
>>
>>> @@ -7931,7 +7928,7 @@ static inline void update_sg_lb_stats(struct
>>> lb_env *env,
>>> if ((env->flags & LBF_NOHZ_STATS) &&
>>> update_nohz_stats(rq, false))
>>> env->flags |= LBF_NOHZ_AGAIN;
>>>
>>> - sgs->group_load += weighted_cpuload(rq);
>>> + sgs->group_load += cpu_load(i);
>>> sgs->group_util += cpu_util(i);
>>> sgs->sum_nr_running += rq->cfs.h_nr_running;
>>
>> ... now we end up dereferencing cpu_rq(cpu) 3 times.
>>
>> I guess per-cpu variables are so cheap that we should
>> never notice, but I thought I'd ask anyway while looking
>> over these patches :)
>
> I was going to say CSE should fix that, but then I noticed per_cpu
> contains that hideous RELOC_HIDE() thing and I figure that might
> confuse GCC enough to break that :/
--->8---
>From 25fcbbd9f654f243a70e38b0d59d38eb3c3f9313 Mon Sep 17 00:00:00 2001
From: Dietmar Eggemann <dietmar.eggem...@arm.com>
Date: Mon, 13 May 2019 11:50:32 +0100
Subject: [PATCH] sched/fair: Rename weighted_cpuload() to cpu_runnable_load()
The term 'weighted' is not needed since there is no 'unweighted' load.
Instead use the term 'runnable' to distinguish 'runnable' load
(avg.runnable_load_avg) used in load balance from load (avg.load_avg)
which is the sum of 'runnable' and 'blocked' load.
Signed-off-by: Dietmar Eggemann <dietmar.eggem...@arm.com>
---
Related to the question whether replacing the 'struct rq *rq' parameter
with 'int cpu' (cpu_rq(cpu)) for cpu_runnable_load() has an influence
on the code:
RELOC_HIDE() (in per_cpu_ptr() -> SHIFT_PERCPU_PTR()) hinders the
compiler to generate similar code (e.g. in update_sg_lb_stats()).
When using 'int cpu' the addressing mode changed from Based Addressing
to Based_indexed-Scaled. Moreover, the text size of fair.o grows by 32
bytes.
kernel/sched/fair.c | 42 +++++++++++++++++++++---------------------
1 file changed, 21 insertions(+), 21 deletions(-)
diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
index 3c11dcdedcbc..0436f8eba556 100644
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -1466,7 +1466,7 @@ bool should_numa_migrate_memory(struct task_struct *p,
struct page * page,
group_faults_cpu(ng, src_nid) * group_faults(p, dst_nid) * 4;
}
-static unsigned long weighted_cpuload(struct rq *rq);
+static unsigned long cpu_runnable_load(struct rq *rq);
/* Cached statistics for all CPUs within a node */
struct numa_stats {
@@ -1487,7 +1487,7 @@ static void update_numa_stats(struct numa_stats *ns, int
nid)
for_each_cpu(cpu, cpumask_of_node(nid)) {
struct rq *rq = cpu_rq(cpu);
- ns->load += weighted_cpuload(rq);
+ ns->load += cpu_runnable_load(rq);
ns->compute_capacity += capacity_of(cpu);
}
@@ -5338,7 +5338,7 @@ static struct {
#endif /* CONFIG_NO_HZ_COMMON */
-static unsigned long weighted_cpuload(struct rq *rq)
+static unsigned long cpu_runnable_load(struct rq *rq)
{
return cfs_rq_runnable_load_avg(&rq->cfs);
}
@@ -5352,7 +5352,7 @@ static unsigned long cpu_avg_load_per_task(int cpu)
{
struct rq *rq = cpu_rq(cpu);
unsigned long nr_running = READ_ONCE(rq->cfs.h_nr_running);
- unsigned long load_avg = weighted_cpuload(rq);
+ unsigned long load_avg = cpu_runnable_load(rq);
if (nr_running)
return load_avg / nr_running;
@@ -5450,7 +5450,7 @@ wake_affine_weight(struct sched_domain *sd, struct
task_struct *p,
s64 this_eff_load, prev_eff_load;
unsigned long task_load;
- this_eff_load = weighted_cpuload(cpu_rq(this_cpu));
+ this_eff_load = cpu_runnable_load(cpu_rq(this_cpu));
if (sync) {
unsigned long current_load = task_h_load(current);
@@ -5468,7 +5468,7 @@ wake_affine_weight(struct sched_domain *sd, struct
task_struct *p,
this_eff_load *= 100;
this_eff_load *= capacity_of(prev_cpu);
- prev_eff_load = weighted_cpuload(cpu_rq(prev_cpu));
+ prev_eff_load = cpu_runnable_load(cpu_rq(prev_cpu));
prev_eff_load -= task_load;
if (sched_feat(WA_BIAS))
prev_eff_load *= 100 + (sd->imbalance_pct - 100) / 2;
@@ -5556,7 +5556,7 @@ find_idlest_group(struct sched_domain *sd, struct
task_struct *p,
max_spare_cap = 0;
for_each_cpu(i, sched_group_span(group)) {
- load = weighted_cpuload(cpu_rq(i));
+ load = cpu_runnable_load(cpu_rq(i));
runnable_load += load;
avg_load += cfs_rq_load_avg(&cpu_rq(i)->cfs);
@@ -5692,7 +5692,7 @@ find_idlest_group_cpu(struct sched_group *group, struct
task_struct *p, int this
shallowest_idle_cpu = i;
}
} else if (shallowest_idle_cpu == -1) {
- load = weighted_cpuload(cpu_rq(i));
+ load = cpu_runnable_load(cpu_rq(i));
if (load < min_load) {
min_load = load;
least_loaded_cpu = i;
@@ -7263,7 +7263,7 @@ static struct task_struct *detach_one_task(struct lb_env
*env)
static const unsigned int sched_nr_migrate_break = 32;
/*
- * detach_tasks() -- tries to detach up to imbalance weighted load from
+ * detach_tasks() -- tries to detach up to imbalance runnable load from
* busiest_rq, as part of a balancing operation within domain "sd".
*
* Returns number of detached tasks if successful and 0 otherwise.
@@ -7331,7 +7331,7 @@ static int detach_tasks(struct lb_env *env)
/*
* We only want to steal up to the prescribed amount of
- * weighted load.
+ * runnable load.
*/
if (env->imbalance <= 0)
break;
@@ -7941,7 +7941,7 @@ static inline void update_sg_lb_stats(struct lb_env *env,
if ((env->flags & LBF_NOHZ_STATS) && update_nohz_stats(rq,
false))
env->flags |= LBF_NOHZ_AGAIN;
- sgs->group_load += weighted_cpuload(rq);
+ sgs->group_load += cpu_runnable_load(rq);
sgs->group_util += cpu_util(i);
sgs->sum_nr_running += rq->cfs.h_nr_running;
@@ -8395,7 +8395,7 @@ static inline void calculate_imbalance(struct lb_env
*env, struct sd_lb_stats *s
* find_busiest_group - Returns the busiest group within the sched_domain
* if there is an imbalance.
*
- * Also calculates the amount of weighted load which should be moved
+ * Also calculates the amount of runnable load which should be moved
* to restore balance.
*
* @env: The load balancing environment.
@@ -8514,7 +8514,7 @@ static struct rq *find_busiest_queue(struct lb_env *env,
int i;
for_each_cpu_and(i, sched_group_span(group), env->cpus) {
- unsigned long capacity, wl;
+ unsigned long capacity, load;
enum fbq_type rt;
rq = cpu_rq(i);
@@ -8568,30 +8568,30 @@ static struct rq *find_busiest_queue(struct lb_env *env,
rq->nr_running == 1)
continue;
- wl = weighted_cpuload(rq);
+ load = cpu_runnable_load(rq);
/*
- * When comparing with imbalance, use weighted_cpuload()
+ * When comparing with imbalance, use cpu_runnable_load()
* which is not scaled with the CPU capacity.
*/
- if (rq->nr_running == 1 && wl > env->imbalance &&
+ if (rq->nr_running == 1 && load > env->imbalance &&
!check_cpu_capacity(rq, env->sd))
continue;
/*
* For the load comparisons with the other CPU's, consider
- * the weighted_cpuload() scaled with the CPU capacity, so
+ * the cpu_runnable_load() scaled with the CPU capacity, so
* that the load can be moved away from the CPU that is
* potentially running at a lower capacity.
*
- * Thus we're looking for max(wl_i / capacity_i), crosswise
+ * Thus we're looking for max(load_i / capacity_i), crosswise
* multiplication to rid ourselves of the division works out
- * to: wl_i * capacity_j > wl_j * capacity_i; where j is
+ * to: load_i * capacity_j > load_j * capacity_i; where j is
* our previous maximum.
*/
- if (wl * busiest_capacity > busiest_load * capacity) {
- busiest_load = wl;
+ if (load * busiest_capacity > busiest_load * capacity) {
+ busiest_load = load;
busiest_capacity = capacity;
busiest = rq;
}
--
2.17.1
