Re: [PATCH] drm/sun4i: sun8i: Add support for pixel blend mode property

From: Roman Stratiienko
Date: Mon Jun 06 2022 - 07:22:09 EST


Hi Maxime,

пн, 6 июн. 2022 г. в 12:22, Maxime Ripard <maxime@xxxxxxxxxx>:
>
> On Mon, Jun 06, 2022 at 11:20:06AM +0200, Maxime Ripard wrote:
> > On Mon, Jun 06, 2022 at 11:17:20AM +0300, Roman Stratiienko wrote:
> > > Hello Jernej,
> > >
> > > Thank you for having a look.
> > >
> > > вс, 5 июн. 2022 г. в 23:37, Jernej Škrabec <jernej.skrabec@xxxxxxxxx>:
> > > >
> > > > Dne nedelja, 05. junij 2022 ob 17:47:31 CEST je Roman Stratiienko napisal(a):
> > > > > Allwinner DE2 and DE3 hardware support 3 pixel blend modes:
> > > > > "None", "Pre-multiplied", "Coverage"
> > > > >
> > > > > Add the blend mode property and route it to the appropriate registers.
> > > > >
> > > > > Note:
> > > > > "force_premulti" parameter was added to handle multi-overlay channel
> > > > > cases in future changes. It must be set to true for cases when more
> > > > > than 1 overlay layer is used within a channel and at least one of the
> > > > > overlay layers within a group uses premultiplied blending mode.
> > > >
> > > > Please remove this parameter. It's nothing special, so it can be easily added
> > > > once it's actually needed. For now, it only complicates code.
> > >
> > > I would prefer keeping it if you do not have any strong opinion against it.
> > >
> > > I am working now on exposing all overlays, so it will be needed soon anyway.
> >
> > KMS assumes pre-multiplied alpha anyway, so we shouldn't need a
> > parameter to force it: we're always going to force it.
>
> My bad, I got confused with your other patch.
>
> Still, I agree with Jernej, if it's not needed it only complicates the
> code for no particular reason. If you need it at some point in the
> future, add it then. Otherwise, it's hard to reason about, since we
> don't know what are the expectations that those future patches will
> bring.

Well, existing code already has some sort of support for the sub-overlays:
For example 'int overlay, ' parameter is always 0, but it still passed through
the call stack. Therefore this patch is just aligned with already existing
traditions to keep sub-overlays in mind.

>
> Maxime

Regards,
Roman