diff mbox series

cpufreq: dt-platdev: mark RK3399 as having separate policies per cluster

Message ID 20181001202153.GA240123@dtor-ws (mailing list archive)
State Changes Requested, archived
Headers show
Series cpufreq: dt-platdev: mark RK3399 as having separate policies per cluster | expand

Commit Message

Dmitry Torokhov Oct. 1, 2018, 8:21 p.m. UTC
RK3399 has one cluster with 4 small cores, and another one with 2 big
cores, with cores in different clusters having different OPPs and thus
different policies. Let's enable this via "have_governor_per_policy"
platform data.

Signed-off-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>
---

Not tested, but we had a patch unconditionally enabling
CPUFREQ_HAVE_GOVERNOR_PER_POLICY flag in tree we used to ship devices
based on RK3399 platform. 

 drivers/cpufreq/cpufreq-dt-platdev.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

Comments

Viresh Kumar Oct. 3, 2018, 4:41 a.m. UTC | #1
On 01-10-18, 13:21, Dmitry Torokhov wrote:
> RK3399 has one cluster with 4 small cores, and another one with 2 big
> cores, with cores in different clusters having different OPPs and thus
> different policies. Let's enable this via "have_governor_per_policy"
> platform data.

The policies are always different in such cases, with or without this flag. What
this flag rather enables is for us to have separate set of tunables for the
governor for individual policies.

For example, without this flag there will be a single governor directory:

/sys/devices/system/cpu/cpufreq/schedutil/

and the value of tunables in that directory will be used for all cpufreq
policies.

With this flag you wouldn't have the governor directory there, but rather in:

/sys/devices/system/cpu/cpufreq/policy*/schedutil/

i.e. tunables per policy and that's why the name of the flag is:
have_governor_per_policy.

> Signed-off-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>
> ---
> 
> Not tested, but we had a patch unconditionally enabling
> CPUFREQ_HAVE_GOVERNOR_PER_POLICY flag in tree we used to ship devices
> based on RK3399 platform. 

Sure, that sounds good. Just that you need to update the commit log a bit.

>  drivers/cpufreq/cpufreq-dt-platdev.c | 5 ++++-
>  1 file changed, 4 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/cpufreq/cpufreq-dt-platdev.c b/drivers/cpufreq/cpufreq-dt-platdev.c
> index fe14c57de6ca..040ec0f711f9 100644
> --- a/drivers/cpufreq/cpufreq-dt-platdev.c
> +++ b/drivers/cpufreq/cpufreq-dt-platdev.c
> @@ -78,7 +78,10 @@ static const struct of_device_id whitelist[] __initconst = {
>  	{ .compatible = "rockchip,rk3328", },
>  	{ .compatible = "rockchip,rk3366", },
>  	{ .compatible = "rockchip,rk3368", },
> -	{ .compatible = "rockchip,rk3399", },
> +	{ .compatible = "rockchip,rk3399",
> +	  .data = &(struct cpufreq_dt_platform_data)

data is void *. No need to provide typecast information. This shouldn't throw
any build warnings I believe.

> +		{ .have_governor_per_policy = true, },
Dmitry Torokhov Oct. 3, 2018, 5:22 a.m. UTC | #2
On Tue, Oct 2, 2018 at 9:41 PM Viresh Kumar <viresh.kumar@linaro.org> wrote:
>
> On 01-10-18, 13:21, Dmitry Torokhov wrote:
> > RK3399 has one cluster with 4 small cores, and another one with 2 big
> > cores, with cores in different clusters having different OPPs and thus
> > different policies. Let's enable this via "have_governor_per_policy"
> > platform data.
>
> The policies are always different in such cases, with or without this flag. What
> this flag rather enables is for us to have separate set of tunables for the
> governor for individual policies.
>
> For example, without this flag there will be a single governor directory:
>
> /sys/devices/system/cpu/cpufreq/schedutil/
>
> and the value of tunables in that directory will be used for all cpufreq
> policies.
>
> With this flag you wouldn't have the governor directory there, but rather in:
>
> /sys/devices/system/cpu/cpufreq/policy*/schedutil/
>
> i.e. tunables per policy and that's why the name of the flag is:
> have_governor_per_policy.
>
> > Signed-off-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>
> > ---
> >
> > Not tested, but we had a patch unconditionally enabling
> > CPUFREQ_HAVE_GOVERNOR_PER_POLICY flag in tree we used to ship devices
> > based on RK3399 platform.
>
> Sure, that sounds good. Just that you need to update the commit log a bit.

OK, I'll do that.

>
> >  drivers/cpufreq/cpufreq-dt-platdev.c | 5 ++++-
> >  1 file changed, 4 insertions(+), 1 deletion(-)
> >
> > diff --git a/drivers/cpufreq/cpufreq-dt-platdev.c b/drivers/cpufreq/cpufreq-dt-platdev.c
> > index fe14c57de6ca..040ec0f711f9 100644
> > --- a/drivers/cpufreq/cpufreq-dt-platdev.c
> > +++ b/drivers/cpufreq/cpufreq-dt-platdev.c
> > @@ -78,7 +78,10 @@ static const struct of_device_id whitelist[] __initconst = {
> >       { .compatible = "rockchip,rk3328", },
> >       { .compatible = "rockchip,rk3366", },
> >       { .compatible = "rockchip,rk3368", },
> > -     { .compatible = "rockchip,rk3399", },
> > +     { .compatible = "rockchip,rk3399",
> > +       .data = &(struct cpufreq_dt_platform_data)
>
> data is void *. No need to provide typecast information. This shouldn't throw
> any build warnings I believe.

We however need to tell the compiler the type of structure we are
creating. The following won't compile:

        .data = &{ .have_governor_per_policy = true, }

Thanks.
diff mbox series

Patch

diff --git a/drivers/cpufreq/cpufreq-dt-platdev.c b/drivers/cpufreq/cpufreq-dt-platdev.c
index fe14c57de6ca..040ec0f711f9 100644
--- a/drivers/cpufreq/cpufreq-dt-platdev.c
+++ b/drivers/cpufreq/cpufreq-dt-platdev.c
@@ -78,7 +78,10 @@  static const struct of_device_id whitelist[] __initconst = {
 	{ .compatible = "rockchip,rk3328", },
 	{ .compatible = "rockchip,rk3366", },
 	{ .compatible = "rockchip,rk3368", },
-	{ .compatible = "rockchip,rk3399", },
+	{ .compatible = "rockchip,rk3399",
+	  .data = &(struct cpufreq_dt_platform_data)
+		{ .have_governor_per_policy = true, },
+	},
 
 	{ .compatible = "st-ericsson,u8500", },
 	{ .compatible = "st-ericsson,u8540", },