Hi
I think so far with my trials port clock changes in sync with phys. As pmdemand atomic check is called many times, I don't think optimizing is bad.
Br
Vinod
From: Sousa, Gustavo <gustavo.sousa@xxxxxxxxx>
Sent: Thursday, 25 May 2023, 17:58
To: Govindapillai, Vinod <vinod.govindapillai@xxxxxxxxx>; intel-gfx@xxxxxxxxxxxxxxxxxxxxx <intel-gfx@xxxxxxxxxxxxxxxxxxxxx>
Cc: Syrjala, Ville <ville.syrjala@xxxxxxxxx>; Lisovskiy, Stanislav <stanislav.lisovskiy@xxxxxxxxx>; Kahola, Mika <mika.kahola@xxxxxxxxx>; Saarinen, Jani <jani.saarinen@xxxxxxxxx>
Subject: Re: [PATCH v8 7/7] drm/i915/mtl: Add support for PM DEMAND
Quoting Govindapillai, Vinod (2023-05-25 11:31:05-03:00)
> Hi
>
> (...)
>
> Can only the phys change without impacting other parameters here?
Not sure to be honest, but, if so, I think we would covering cases like
that with the proposed change. IMO, if we depend on PHYs being
active/non-active, we should track that to in pmdemand_needs_update() be
safe.
One scenario I could think of is an update to switch connectors while
keeping current pipe configuration.
--
Gustavo Sousa
