[PATCH] sched/topology: One function call less in build_group_from_child_sched_domain()

[PATCH] sched/topology: One function call less in build_group_from_child_sched_domain()

[Markus Elfring]

From: Markus Elfring <elfring@users.sourceforge.net>
Date: Sat, 6 Jul 2019 16:00:13 +0200

Avoid an extra function call by using a ternary operator instead of
a conditional statement.

This issue was detected by using the Coccinelle software.

Signed-off-by: Markus Elfring <elfring@users.sourceforge.net>
---
 kernel/sched/topology.c | 6 +-----
 1 file changed, 1 insertion(+), 5 deletions(-)

diff --git a/kernel/sched/topology.c b/kernel/sched/topology.c
index f751ce0b783e..6190eb52c30a 100644
--- a/kernel/sched/topology.c
+++ b/kernel/sched/topology.c
@@ -886,11 +886,7 @@ build_group_from_child_sched_domain(struct sched_domain *sd, int cpu)
 		return NULL;

 	sg_span = sched_group_span(sg);
-	if (sd->child)
-		cpumask_copy(sg_span, sched_domain_span(sd->child));
-	else
-		cpumask_copy(sg_span, sched_domain_span(sd));
-
+	cpumask_copy(sg_span, sched_domain_span(sd->child ? sd->child : sd));
 	atomic_inc(&sg->ref);
 	return sg;
 }
--
2.22.0

Re: [PATCH] sched/topology: One function call less in build_group_from_child_sched_domain()

[Srikar Dronamraju]

* Markus Elfring <Markus.Elfring@web.de> [2019-07-06 16:05:17]:

> From: Markus Elfring <elfring@users.sourceforge.net>
> Date: Sat, 6 Jul 2019 16:00:13 +0200
> 
> Avoid an extra function call by using a ternary operator instead of
> a conditional statement.
> 
> This issue was detected by using the Coccinelle software.
> 
> Signed-off-by: Markus Elfring <elfring@users.sourceforge.net>
> ---
>  kernel/sched/topology.c | 6 +-----
>  1 file changed, 1 insertion(+), 5 deletions(-)
> 
> diff --git a/kernel/sched/topology.c b/kernel/sched/topology.c
> index f751ce0b783e..6190eb52c30a 100644
> --- a/kernel/sched/topology.c
> +++ b/kernel/sched/topology.c
> @@ -886,11 +886,7 @@ build_group_from_child_sched_domain(struct sched_domain *sd, int cpu)
>  		return NULL;
> 
>  	sg_span = sched_group_span(sg);
> -	if (sd->child)
> -		cpumask_copy(sg_span, sched_domain_span(sd->child));
> -	else
> -		cpumask_copy(sg_span, sched_domain_span(sd));
> -
> +	cpumask_copy(sg_span, sched_domain_span(sd->child ? sd->child : sd));

At runtime, Are we avoiding a function call?
However I think we are avoiding a branch instead of a conditional, which may
be beneficial.

>  	atomic_inc(&sg->ref);
>  	return sg;
>  }
> --
> 2.22.0
> 

-- 
Thanks and Regards
Srikar Dronamraju

Re: [PATCH] sched/topology: One function call less in build_group_from_child_sched_domain()

[Enrico Weigelt, metux IT consult]

On 06.07.19 19:22, Srikar Dronamraju wrote:
> * Markus Elfring <Markus.Elfring@web.de> [2019-07-06 16:05:17]:
> 
>> From: Markus Elfring <elfring@users.sourceforge.net>
>> Date: Sat, 6 Jul 2019 16:00:13 +0200
>>
>> Avoid an extra function call by using a ternary operator instead of
>> a conditional statement.
>>
>> This issue was detected by using the Coccinelle software.
>>
>> Signed-off-by: Markus Elfring <elfring@users.sourceforge.net>
>> ---
>>  kernel/sched/topology.c | 6 +-----
>>  1 file changed, 1 insertion(+), 5 deletions(-)
>>
>> diff --git a/kernel/sched/topology.c b/kernel/sched/topology.c
>> index f751ce0b783e..6190eb52c30a 100644
>> --- a/kernel/sched/topology.c
>> +++ b/kernel/sched/topology.c
>> @@ -886,11 +886,7 @@ build_group_from_child_sched_domain(struct sched_domain *sd, int cpu)
>>  		return NULL;
>>
>>  	sg_span = sched_group_span(sg);
>> -	if (sd->child)
>> -		cpumask_copy(sg_span, sched_domain_span(sd->child));
>> -	else
>> -		cpumask_copy(sg_span, sched_domain_span(sd));
>> -
>> +	cpumask_copy(sg_span, sched_domain_span(sd->child ? sd->child : sd));
> 
> At runtime, Are we avoiding a function call?
> However I think we are avoiding a branch instead of a conditional, which may
> be beneficial.

If you're assuming the compiler doesn't already optimize that (no idea
whether gcc really does that).

@Markus: could you check what gcc is actually generating out of both the
old and your new version ?


--mtx

-- 
Enrico Weigelt, metux IT consult
Free software and Linux embedded engineering
info@metux.net -- +49-151-27565287

Re: [PATCH] sched/topology: One function call less in build_group_from_child_sched_domain()

[Srikar Dronamraju]

* Enrico Weigelt, metux IT consult <lkml@metux.net> [2019-07-08 11:38:58]:

> > 
> > At runtime, Are we avoiding a function call?
> > However I think we are avoiding a branch instead of a conditional, which may
> > be beneficial.
> 
> If you're assuming the compiler doesn't already optimize that (no idea
> whether gcc really does that).
> 
> @Markus: could you check what gcc is actually generating out of both the
> old and your new version ?
> 
> 

I had already tried looking at the object files both on X86 and PowerPc and
in both cases (with and without patch) the generated code differs.

> --mtx
> 
> -- 
> Enrico Weigelt, metux IT consult
> Free software and Linux embedded engineering
> info@metux.net -- +49-151-27565287
> 

-- 
Thanks and Regards
Srikar Dronamraju

Re: [PATCH] sched/topology: One function call less in build_group_from_child_sched_domain()

[Peter Zijlstra]

On Sat, Jul 06, 2019 at 10:52:23PM +0530, Srikar Dronamraju wrote:
> * Markus Elfring <Markus.Elfring@web.de> [2019-07-06 16:05:17]:
> 
> > From: Markus Elfring <elfring@users.sourceforge.net>
> > Date: Sat, 6 Jul 2019 16:00:13 +0200
> > 
> > Avoid an extra function call by using a ternary operator instead of
> > a conditional statement.
> > 
> > This issue was detected by using the Coccinelle software.
> > 
> > Signed-off-by: Markus Elfring <elfring@users.sourceforge.net>
> > ---
> >  kernel/sched/topology.c | 6 +-----
> >  1 file changed, 1 insertion(+), 5 deletions(-)
> > 
> > diff --git a/kernel/sched/topology.c b/kernel/sched/topology.c
> > index f751ce0b783e..6190eb52c30a 100644
> > --- a/kernel/sched/topology.c
> > +++ b/kernel/sched/topology.c
> > @@ -886,11 +886,7 @@ build_group_from_child_sched_domain(struct sched_domain *sd, int cpu)
> >  		return NULL;
> > 
> >  	sg_span = sched_group_span(sg);
> > -	if (sd->child)
> > -		cpumask_copy(sg_span, sched_domain_span(sd->child));
> > -	else
> > -		cpumask_copy(sg_span, sched_domain_span(sd));
> > -
> > +	cpumask_copy(sg_span, sched_domain_span(sd->child ? sd->child : sd));
> 
> At runtime, Are we avoiding a function call?
> However I think we are avoiding a branch instead of a conditional, which may
> be beneficial.

It all depends on what the compiler does; also this is super slow path
stuff and the patch makes code less readable (IMO).

Re: [PATCH] sched/topology: One function call less in build_group_from_child_sched_domain()

[Srikar Dronamraju]

* Peter Zijlstra <peterz@infradead.org> [2019-07-08 12:23:12]:

> On Sat, Jul 06, 2019 at 10:52:23PM +0530, Srikar Dronamraju wrote:
> > * Markus Elfring <Markus.Elfring@web.de> [2019-07-06 16:05:17]:
> > 
> > > From: Markus Elfring <elfring@users.sourceforge.net>
> > > Date: Sat, 6 Jul 2019 16:00:13 +0200
> > > 
> > > Avoid an extra function call by using a ternary operator instead of
> > > a conditional statement.
> > > 
> > > This issue was detected by using the Coccinelle software.
> > > 
> > > Signed-off-by: Markus Elfring <elfring@users.sourceforge.net>
> > > ---
> > >  kernel/sched/topology.c | 6 +-----
> > >  1 file changed, 1 insertion(+), 5 deletions(-)
> > > 
> > > diff --git a/kernel/sched/topology.c b/kernel/sched/topology.c
> > > index f751ce0b783e..6190eb52c30a 100644
> > > --- a/kernel/sched/topology.c
> > > +++ b/kernel/sched/topology.c
> > > @@ -886,11 +886,7 @@ build_group_from_child_sched_domain(struct sched_domain *sd, int cpu)
> > >  		return NULL;
> > > 
> > >  	sg_span = sched_group_span(sg);
> > > -	if (sd->child)
> > > -		cpumask_copy(sg_span, sched_domain_span(sd->child));
> > > -	else
> > > -		cpumask_copy(sg_span, sched_domain_span(sd));
> > > -
> > > +	cpumask_copy(sg_span, sched_domain_span(sd->child ? sd->child : sd));
> > 
> > At runtime, Are we avoiding a function call?
> > However I think we are avoiding a branch instead of a conditional, which may
> > be beneficial.
> 
> It all depends on what the compiler does; also this is super slow path
> stuff and the patch makes code less readable (IMO).
> 

Yes, it definitely makes code readable. I was only commenting on the
changelog/subject which says avoids a function call which I think it
doesn't.

-- 
Thanks and Regards
Srikar Dronamraju

Re: sched/topology: One function call less in build_group_from_child_sched_domain()

[Markus Elfring]

> Yes, it definitely makes code readable.

Some developers seem to get difficulties with this view.


> I was only commenting on the changelog/subject which says
> avoids a function call which I think it doesn't.

Do you prefer an other wording for calling functions only once
with a selected parameter value instead of calling them (unoptimised)
within two branches of an if statement?

Regards,
Markus