cpumask_weight() is known to be expensive. In Credit2, we use it in
load-balancing, but only for knowing how many CPUs are active in a
runqueue.
Keeping such count in an integer field of the per-runqueue data
structure we have, completely avoids the need for cpumask_weight().
While there, remove as much other uses of it as we can, even if not in
hot-paths.
Signed-off-by: Dario Faggioli <dfaggioli@suse.com>
---
Cc: George Dunlap <george.dunlap@eu.citrix.com>
Cc: Andrew Cooper <andrew.cooper3@citrix.com>
Cc: Andrii Anisov <andrii.anisov@gmail.com>
---
I just realized this patch fell through the cracks! :-O
I sent it quite a while ago, I got some comments from Andrew and
Andrii, but then never really followed up.
So here we hare!
Changes from v1:
* nr_cpus is now unsigned
* coding style (no hard tabs)
* killed a couple of redundant ASSERT()s
---
xen/common/sched_credit2.c | 19 +++++++++++++------
1 file changed, 13 insertions(+), 6 deletions(-)
diff --git a/xen/common/sched_credit2.c b/xen/common/sched_credit2.c
index 8e4381d8a7..fbdc4618cb 100644
--- a/xen/common/sched_credit2.c
+++ b/xen/common/sched_credit2.c
@@ -466,6 +466,7 @@ struct csched2_runqueue_data {
spinlock_t lock; /* Lock for this runqueue */
struct list_head runq; /* Ordered list of runnable vms */
+ unsigned int nr_cpus; /* How many CPUs are sharing this runqueue */
int id; /* ID of this runqueue (-1 if invalid) */
int load; /* Instantaneous load (num of non-idle vcpus) */
@@ -2613,8 +2614,8 @@ retry:
if ( st.orqd->b_avgload > load_max )
load_max = st.orqd->b_avgload;
- cpus_max = cpumask_weight(&st.lrqd->active);
- i = cpumask_weight(&st.orqd->active);
+ cpus_max = st.lrqd->nr_cpus;
+ i = st.orqd->nr_cpus;
if ( i > cpus_max )
cpus_max = i;
@@ -3697,7 +3698,7 @@ csched2_dump(const struct scheduler *ops)
"\tinstload = %d\n"
"\taveload = %"PRI_stime" (~%"PRI_stime"%%)\n",
i,
- cpumask_weight(&prv->rqd[i].active),
+ prv->rqd[i].nr_cpus,
nr_cpu_ids, cpumask_bits(&prv->rqd[i].active),
prv->rqd[i].max_weight,
prv->rqd[i].pick_bias,
@@ -3815,7 +3816,7 @@ init_pdata(struct csched2_private *prv, struct csched2_pcpu *spc,
__cpumask_set_cpu(cpu, &spc->sibling_mask);
- if ( cpumask_weight(&rqd->active) > 0 )
+ if ( rqd->nr_cpus > 0 )
for_each_cpu ( rcpu, per_cpu(cpu_sibling_mask, cpu) )
if ( cpumask_test_cpu(rcpu, &rqd->active) )
{
@@ -3828,7 +3829,10 @@ init_pdata(struct csched2_private *prv, struct csched2_pcpu *spc,
__cpumask_set_cpu(cpu, &prv->initialized);
__cpumask_set_cpu(cpu, &rqd->smt_idle);
- if ( cpumask_weight(&rqd->active) == 1 )
+ rqd->nr_cpus++;
+ ASSERT(cpumask_weight(&rqd->active) == rqd->nr_cpus);
+
+ if ( rqd->nr_cpus == 1 )
rqd->pick_bias = cpu;
return spc->runq_id;
@@ -3934,7 +3938,10 @@ csched2_deinit_pdata(const struct scheduler *ops, void *pcpu, int cpu)
for_each_cpu ( rcpu, &rqd->active )
__cpumask_clear_cpu(cpu, &csched2_pcpu(rcpu)->sibling_mask);
- if ( cpumask_empty(&rqd->active) )
+ rqd->nr_cpus--;
+ ASSERT(cpumask_weight(&rqd->active) == rqd->nr_cpus);
+
+ if ( rqd->nr_cpus == 0 )
{
printk(XENLOG_INFO " No cpus left on runqueue, disabling\n");
deactivate_runqueue(prv, spc->runq_id);
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xenproject.org
https://lists.xenproject.org/mailman/listinfo/xen-devel
On 29.07.19 13:49, Dario Faggioli wrote: > cpumask_weight() is known to be expensive. In Credit2, we use it in > load-balancing, but only for knowing how many CPUs are active in a > runqueue. > > Keeping such count in an integer field of the per-runqueue data > structure we have, completely avoids the need for cpumask_weight(). > > While there, remove as much other uses of it as we can, even if not in > hot-paths. > > Signed-off-by: Dario Faggioli <dfaggioli@suse.com> > --- > Cc: George Dunlap <george.dunlap@eu.citrix.com> > Cc: Andrew Cooper <andrew.cooper3@citrix.com> > Cc: Andrii Anisov <andrii.anisov@gmail.com> > --- > I just realized this patch fell through the cracks! :-O > > I sent it quite a while ago, I got some comments from Andrew and > Andrii, but then never really followed up. > > So here we hare! > > Changes from v1: > * nr_cpus is now unsigned > * coding style (no hard tabs) > * killed a couple of redundant ASSERT()s > --- Reviewed-by: Andrii Anisov <andrii_anisov@epam.com> -- Sincerely, Andrii Anisov. _______________________________________________ Xen-devel mailing list Xen-devel@lists.xenproject.org https://lists.xenproject.org/mailman/listinfo/xen-devel
On 7/29/19 11:49 AM, Dario Faggioli wrote: > cpumask_weight() is known to be expensive. In Credit2, we use it in > load-balancing, but only for knowing how many CPUs are active in a > runqueue. > > Keeping such count in an integer field of the per-runqueue data > structure we have, completely avoids the need for cpumask_weight(). > > While there, remove as much other uses of it as we can, even if not in > hot-paths. > > Signed-off-by: Dario Faggioli <dfaggioli@suse.com> Looks good, thanks. Reviewed-by: George Dunlap <george.dunlap@citrix.com> _______________________________________________ Xen-devel mailing list Xen-devel@lists.xenproject.org https://lists.xenproject.org/mailman/listinfo/xen-devel
© 2016 - 2024 Red Hat, Inc.