[QUESTION] drm/meson: Hardware cursor plane for Amlogic/meson

[QUESTION] drm/meson: Hardware cursor plane for Amlogic/meson

[Jian-Hong Pan]

We have a khadas VIM2 board equipped with Amlogic S912 chip. We found
it shows multiple cursors at the same screen which is more like
afterimages. This bug is filed as the ticket [1] and exists until X
uses software cursor with the commit ("modesetting: Disable
pageflipping when using a swcursor") [2]. Meson drm does not support
hardware cursors and current versions of X will disable pageflipping
when using software cursor, which is a big performance hit.

We also learned there is the hardware cursor plane supported by
Amlogic chips from the slide "Mainline Linux on AmLogic SoCs" [3]. Is
it possible to have the hardware cursor available over the drm
interface?

[1] https://gitlab.freedesktop.org/mesa/mesa/-/issues/3143
[2] https://gitlab.freedesktop.org/xorg/xserver/-/commit/0aaac8d783e78c040a70a55ba8d67809abd7e625
[3] https://people.freedesktop.org/~narmstrong/lca2018-amlogic/#/section

Thanks,
Jian-Hong Pan

_______________________________________________
linux-amlogic mailing list
linux-amlogic@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-amlogic

Re: [QUESTION] drm/meson: Hardware cursor plane for Amlogic/meson

[Daniel Drake]

Hi Neil,

On Wed, Aug 5, 2020 at 3:53 PM Jian-Hong Pan <jian-hong@endlessm.com> wrote:
> Meson drm does not support
> hardware cursors and current versions of X will disable pageflipping
> when using software cursor, which is a big performance hit. [2]
>
> We also learned there is the hardware cursor plane supported by
> Amlogic chips from the slide "Mainline Linux on AmLogic SoCs" [3]. Is
> it possible to have the hardware cursor available over the drm
> interface?
>
> [1] https://gitlab.freedesktop.org/mesa/mesa/-/issues/3143
> [2] https://gitlab.freedesktop.org/xorg/xserver/-/commit/0aaac8d783e78c040a70a55ba8d67809abd7e625
> [3] https://people.freedesktop.org/~narmstrong/lca2018-amlogic/#/section

Curious if you have any quick input here. If it's feasible to enable
the drm cursor plane on meson-drm then X would be able to perform a
lot better especially now that we have usable GPU drivers too.

Thanks
Daniel

_______________________________________________
linux-amlogic mailing list
linux-amlogic@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-amlogic

Re: [QUESTION] drm/meson: Hardware cursor plane for Amlogic/meson

[Martin Blumenstingl]

Hello Jian-Hong Pan,

On Wed, Aug 5, 2020 at 9:54 AM Jian-Hong Pan <jian-hong@endlessm.com> wrote:
[...]
> We also learned there is the hardware cursor plane supported by
> Amlogic chips from the slide "Mainline Linux on AmLogic SoCs" [3]. Is
> it possible to have the hardware cursor available over the drm
> interface?
as far as I understand it is technically possible but currently not implemented.
We have a specific TODO item for this, see [0]

also I'm not sure if this is trivial to implement and whether there
are any plans to do so


Best regards,
Martin


[0] http://linux-meson.com/doku.php#todo

_______________________________________________
linux-amlogic mailing list
linux-amlogic@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-amlogic

Re: [QUESTION] drm/meson: Hardware cursor plane for Amlogic/meson

[Neil Armstrong]

Hi,

On 15/08/2020 10:59, Martin Blumenstingl wrote:
> Hello Jian-Hong Pan,
> 
> On Wed, Aug 5, 2020 at 9:54 AM Jian-Hong Pan <jian-hong@endlessm.com> wrote:
> [...]
>> We also learned there is the hardware cursor plane supported by
>> Amlogic chips from the slide "Mainline Linux on AmLogic SoCs" [3]. Is
>> it possible to have the hardware cursor available over the drm
>> interface?
> as far as I understand it is technically possible but currently not implemented.
> We have a specific TODO item for this, see [0]
> 
> also I'm not sure if this is trivial to implement and whether there
> are any plans to do so

Sorry I was on vacation !

Yes it's technically possible, but not trivial.

The blocking point is Interlaced support, currently we use the single available
scaler to automatically switch the field on each vsync on the primary plane.

But if we were using a secondary plane for cursor, we would need to switch the field
manually for this plane.

Clearly it's not an issue, but the plane management has changed for G12A and support for
this would need to be added at the same time.

We (BayLibre) do not plan to implement it in the short or long term, but it's clearly a wanted feature.

Neil

> 
> 
> Best regards,
> Martin
> 
> 
> [0] http://linux-meson.com/doku.php#todo
> 
> _______________________________________________
> linux-amlogic mailing list
> linux-amlogic@lists.infradead.org
> http://lists.infradead.org/mailman/listinfo/linux-amlogic
> 


_______________________________________________
linux-amlogic mailing list
linux-amlogic@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-amlogic