Re: [PATCH v4 0/5] Add Qualcomm JPEG V4L2 encoder for SM8250

[Vladimir Zapolskiy]

On 7/6/26 17:49, Bryan O'Donoghue wrote:
> On 06/07/2026 15:35, Vladimir Zapolskiy wrote:
> > > omitting power-domains for individual nodes is incorrect. Best practice
> > > and in fact the _only_ practice that can work is to list power-domains,
> >
> > Let's step back and discuss hardware description of CAMSS IPs, so far
> > there is no point to jump to the practice.
> >
> > > interconnects in the individual nodes themselves as - I've pointed out
> > > multiple times now - those blocks have to ramp clocks and scale voltages
> > > dependent on their particular use cases.
> >
> > Right, you've pointed it a few times, no surprise a straw man fallacy
> > argument resembles as is, because it does not address or debate the
> > initial point of concern [1], when the absolutely excessive complexity
> > is proposed to be implemented for "CAMSS power island" and its IPs.
> >
> > [1] https://lore.kernel.org/linux-media/e04cb9dc-76dd-4fa8-92f4-
> > be002bf7ee8b@xxxxxxxxxx
>
> You seem a little stuck on the term "CAMSS power island" the individual
> blocks within the CAMSS "island" are individually collapsible.

It's not me, who stuck, but dt bindings documentation describes CAMSS on
a number of Qualcomm platforms.

> That is why they _must_ have their own power-domains.
>
> The CamNoc is shared between these components but not outside. Hence it
> is more like a bus than a parent/child relationship.
>
> As to your argument about complexity. I think that's subjective.

What is subjective between one instance and multiple identical instances
of hardware properties? The first is simple the latter is complex.

--
Best wishes,
Vladimir