Re: [PATCH 0/8] Reup: SM8350 and SC8280XP venus support

[Dmitry Baryshkov <dmitry.baryshkov@oss.qualcomm.com>]

On 04/04/2025 08:24, Vikash Garodia wrote:
> Hi Dmitry,
> 
> On 4/3/2025 10:28 PM, Dmitry Baryshkov wrote:
>> On Wed, Mar 05, 2025 at 08:49:37AM +0530, Vikash Garodia wrote:
>>>
>>> On 3/4/2025 6:37 PM, Bryan O'Donoghue wrote:
>>>> This series is a re-up of Konrad's original venus series for sc8280xp and
>>>> sm8350.Why this is enabled on venus driver ? Why not iris driver ? This needs an
>>> explanation on was this even tried to bring up on iris driver.
>>>
>>> How different is this from sm8250 which is already enabled on iris driver ?
>>
>> As far as I remember, SM8250 support in Iris did not reach
>> feature-parity yet. So in my opinion it is fine to add new platforms to
>> the Venus driver, that will later migrate to the Iris driver.
> I would say, from decoder side all codecs are there now on Iris. H264 merged,
> while h265 and VP9 dec are posted as RFC, there is one compliance failure which
> is under debug to post them as regular patches.
> If we are mainly looking for decode usecases, then we should be on Iris.

No, we are not limited to the decode use case.

> Preference would be to stay on Iris, otherwise we would have that extra ask to
> port it later from venus to iris.

Yes, but that would (hopefully) be easy to handle.

>>
>> Otherwise users of SC8280XP either have to use external patchsets (like
>> this one) or a non-full-featured driver (and still possibly external
>> patchsets, I didn't check if these two platforms can use
>> qcom,sm8250-venus as a fallback compat string).
> It should, atleast from the hardware spec perspective, AFAIK.
>>
>> Bryan, Konrad, in my opinion, let's get these patches merged :-)
>>
>>>
>>>> Link: https://lore.kernel.org/all/20230731-topic-8280_venus-v1-0-8c8bbe1983a5@linaro.org/
>>>>
>>>> The main obstacle to merging that series at the time was the longstanding
>>>> but invalid usage of "video-encoder" and "video-decoder" which is a
>>>> driver level configuration option not a description of hardware.
>>>>
>>>> Following on from that discussion a backwards compatible means of
>>>> statically selecting transcoder mode was upstreamed
>>>>
>>>> commit: 687bfbba5a1c ("media: venus: Add support for static video encoder/decoder declarations")
>>>>
>>>> Reworking this series from Konrad to incorporate this simple change
>>>>
>>
> Regards,
> Vikash


-- 
With best wishes
Dmitry