2022-01-01 16:40:20

by Abel Vesa

[permalink] [raw]
Subject: [PATCH v2 2/3] interconnect: imx: Add imx_icc_get_bw and imx_icc_aggregate functions

The aggregate function will return whatever is the highest
rate for that specific node. The imx_icc_get_bw sets the
initial avg and peak to 0 in order to avoid setting them to
INT_MAX by the interconnect core.

Signed-off-by: Abel Vesa <[email protected]>
---

No changes since v1.

drivers/interconnect/imx/imx.c | 20 +++++++++++++++++++-
1 file changed, 19 insertions(+), 1 deletion(-)

diff --git a/drivers/interconnect/imx/imx.c b/drivers/interconnect/imx/imx.c
index 34bfc7936387..4d8a2a2d2608 100644
--- a/drivers/interconnect/imx/imx.c
+++ b/drivers/interconnect/imx/imx.c
@@ -25,6 +25,23 @@ struct imx_icc_node {
struct dev_pm_qos_request qos_req;
};

+static int imx_icc_get_bw(struct icc_node *node, u32 *avg, u32 *peak)
+{
+ *avg = 0;
+ *peak = 0;
+
+ return 0;
+}
+
+static int imx_icc_aggregate(struct icc_node *node, u32 tag, u32 avg_bw,
+ u32 peak_bw, u32 *agg_avg, u32 *agg_peak)
+{
+ *agg_avg = max(*agg_avg, avg_bw);
+ *agg_peak = max(*agg_peak, peak_bw);
+
+ return 0;
+}
+
static int imx_icc_node_set(struct icc_node *node)
{
struct device *dev = node->provider->dev;
@@ -233,7 +250,8 @@ int imx_icc_register(struct platform_device *pdev,
if (!provider)
return -ENOMEM;
provider->set = imx_icc_set;
- provider->aggregate = icc_std_aggregate;
+ provider->get_bw = imx_icc_get_bw;
+ provider->aggregate = imx_icc_aggregate;
provider->xlate = of_icc_xlate_onecell;
provider->data = data;
provider->dev = dev;
--
2.31.1



2022-01-01 18:26:15

by Georgi Djakov

[permalink] [raw]
Subject: Re: [PATCH v2 2/3] interconnect: imx: Add imx_icc_get_bw and imx_icc_aggregate functions

Hi Abel,

On 1.01.22 18:39, Abel Vesa wrote:
> The aggregate function will return whatever is the highest
> rate for that specific node. The imx_icc_get_bw sets the

Adding some more details about why we switch from
icc_std_aggregate to imx_icc_aggregate would be nice.

> initial avg and peak to 0 in order to avoid setting them to
> INT_MAX by the interconnect core.

Do we need a Fixes tag for this?

I would recommend to split imx_icc_get_bw and imx_icc_aggregate
changes into separate patches. These also seem to be unrelated to
the imx_icc_node_adj_desc patchset.

Thanks,
Georgi

> Signed-off-by: Abel Vesa <[email protected]>
> ---
>
> No changes since v1.
>
> drivers/interconnect/imx/imx.c | 20 +++++++++++++++++++-
> 1 file changed, 19 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/interconnect/imx/imx.c b/drivers/interconnect/imx/imx.c
> index 34bfc7936387..4d8a2a2d2608 100644
> --- a/drivers/interconnect/imx/imx.c
> +++ b/drivers/interconnect/imx/imx.c
> @@ -25,6 +25,23 @@ struct imx_icc_node {
> struct dev_pm_qos_request qos_req;
> };
>
> +static int imx_icc_get_bw(struct icc_node *node, u32 *avg, u32 *peak)
> +{
> + *avg = 0;
> + *peak = 0;
> +
> + return 0;
> +}
> +
> +static int imx_icc_aggregate(struct icc_node *node, u32 tag, u32 avg_bw,
> + u32 peak_bw, u32 *agg_avg, u32 *agg_peak)
> +{
> + *agg_avg = max(*agg_avg, avg_bw);
> + *agg_peak = max(*agg_peak, peak_bw);
> +
> + return 0;
> +}
> +
> static int imx_icc_node_set(struct icc_node *node)
> {
> struct device *dev = node->provider->dev;
> @@ -233,7 +250,8 @@ int imx_icc_register(struct platform_device *pdev,
> if (!provider)
> return -ENOMEM;
> provider->set = imx_icc_set;
> - provider->aggregate = icc_std_aggregate;
> + provider->get_bw = imx_icc_get_bw;
> + provider->aggregate = imx_icc_aggregate;
> provider->xlate = of_icc_xlate_onecell;
> provider->data = data;
> provider->dev = dev;


2022-01-03 15:02:04

by Abel Vesa

[permalink] [raw]
Subject: Re: [PATCH v2 2/3] interconnect: imx: Add imx_icc_get_bw and imx_icc_aggregate functions

On 22-01-01 20:26:05, Georgi Djakov wrote:
> Hi Abel,
>
> On 1.01.22 18:39, Abel Vesa wrote:
> > The aggregate function will return whatever is the highest
> > rate for that specific node. The imx_icc_get_bw sets the
>
> Adding some more details about why we switch from
> icc_std_aggregate to imx_icc_aggregate would be nice.
>

On a second look, I think I can drop the imx_icc_aggregate and use the
icc_std_aggregate instead, as long as I use the peak_bw and forget about
the agg_bw.

> > initial avg and peak to 0 in order to avoid setting them to
> > INT_MAX by the interconnect core.
>
> Do we need a Fixes tag for this?
>

Neah, the imx interconnect is not used by any platform yet.

> I would recommend to split imx_icc_get_bw and imx_icc_aggregate
> changes into separate patches. These also seem to be unrelated to
> the imx_icc_node_adj_desc patchset.
>

Since I can use icc_std_aggregate, the imx_icc_aggregate change will be
dropped.

> Thanks,
> Georgi
>
> > Signed-off-by: Abel Vesa <[email protected]>
> > ---
> >
> > No changes since v1.
> >
> > drivers/interconnect/imx/imx.c | 20 +++++++++++++++++++-
> > 1 file changed, 19 insertions(+), 1 deletion(-)
> >
> > diff --git a/drivers/interconnect/imx/imx.c b/drivers/interconnect/imx/imx.c
> > index 34bfc7936387..4d8a2a2d2608 100644
> > --- a/drivers/interconnect/imx/imx.c
> > +++ b/drivers/interconnect/imx/imx.c
> > @@ -25,6 +25,23 @@ struct imx_icc_node {
> > struct dev_pm_qos_request qos_req;
> > };
> > +static int imx_icc_get_bw(struct icc_node *node, u32 *avg, u32 *peak)
> > +{
> > + *avg = 0;
> > + *peak = 0;
> > +
> > + return 0;
> > +}
> > +
> > +static int imx_icc_aggregate(struct icc_node *node, u32 tag, u32 avg_bw,
> > + u32 peak_bw, u32 *agg_avg, u32 *agg_peak)
> > +{
> > + *agg_avg = max(*agg_avg, avg_bw);
> > + *agg_peak = max(*agg_peak, peak_bw);
> > +
> > + return 0;
> > +}
> > +
> > static int imx_icc_node_set(struct icc_node *node)
> > {
> > struct device *dev = node->provider->dev;
> > @@ -233,7 +250,8 @@ int imx_icc_register(struct platform_device *pdev,
> > if (!provider)
> > return -ENOMEM;
> > provider->set = imx_icc_set;
> > - provider->aggregate = icc_std_aggregate;
> > + provider->get_bw = imx_icc_get_bw;
> > + provider->aggregate = imx_icc_aggregate;
> > provider->xlate = of_icc_xlate_onecell;
> > provider->data = data;
> > provider->dev = dev;
>