Re: [PATCH 2/2] media: iris: add new rate control type MBR for encoder

[Vikash Garodia <vikash.garodia@oss.qualcomm.com>]

On 3/30/2026 11:14 AM, Sachin Kumar Garg wrote:
> 
> 
> On 3/17/2026 12:17 AM, Dmitry Baryshkov wrote:
>> On Mon, Mar 16, 2026 at 01:37:21PM -0400, Nicolas Dufresne wrote:
>>> Le lundi 16 mars 2026 à 11:52 +0530, Sachin Kumar Garg a écrit :
>>>>
>>>>
>>>> On 2/13/2026 3:24 PM, Konrad Dybcio wrote:
>>>>> On 2/13/26 7:04 AM, Sachin Kumar Garg wrote:
>>>>>> Introduce V4L2_MPEG_VIDEO_BITRATE_MODE_MBR rate control to Encoder.
>>>>>> Encoder will choose appropriate quantization parameter and
>>>>>> do the smart bit allocation to set the frame maximum bitrate
>>>>>> level as per the Bitrate value configured.
>>>>>> ---
>>>>>>    drivers/media/platform/qcom/iris/iris_ctrls.c      |   2 +
>>>>>>    .../platform/qcom/iris/iris_hfi_gen1_defines.h     |   1 +
>>>>>>    .../media/platform/qcom/iris/iris_platform_gen1.c  |   9 +-
>>>>>>    .../platform/qcom/iris/iris_platform_sc7280.h      | 202
>>>>>> +++++++++++++++++++++
>>>>>
>>>>> Is this really only available on 7280?
>>>>>
>>>>> Konrad
>>>>
>>>> Yes Konrad, in this series we enable V4L2_MPEG_VIDEO_BITRATE_MODE_MBR
>>>> only for SC7280 because currently, we have added the support only for
>>>> low tier chipset. MBR rate control aims to improve the compression
>>>> efficiency of encoder for static and low motion scenes for extremely 
>>>> low
>>>> bitrate use-case.
>>>
>>> Without a spec definition, which constrained the implementations toward
>>> interoperability, this needs to be made vendor specific. That being 
>>> said, even
>>> as vendor specific control, I think it deserves more documentation 
>>> for your
>>> users.
>>
>> What kind of documentation would be a better fit for keeping this in the
>> generic namespace? Frankly speaking, I think that the current
>> documentation is too vendor-specific, it describe the particular
>> algorithm, while the control seems to be more generic. However, I also
>> feel that defining it as "Maximum Bitrate" might not be specific enough
>> (nor would it show a difference from the CBR mode).
>>
>>
> Thanks for the suggestion.
> I understand that this functionality is vendor-specific and should not 
> be exposed via a generic control. Would it be acceptable to convert this 
> into a vendor-specific control (for example, 
> V4L2_MPEG_VIDEO_BITRATE_MODE_QCOM_MBR) and update the documentation 
> accordingly?
> Please let me know if this aligns with upstream expectations, or can you 
> please suggest if there is any other preferred approach for adding 
> vendor-specific controls in this case.

I discussed this with Hans during embedded conference on handling vendor 
specific controls. His suggestion was that vendor can define a reserve 
for 16/32/64 controls, something similar to [1], for cases like this 
which seems quite vendor specific handling to me.

[1] 
https://github.com/torvalds/linux/blob/0e35b9b6ec0ffcc5e23cbdec09f5c622ad532b53/include/uapi/linux/v4l2-controls.h#L235

Regards,
Vikash

> 
> Thanks,
> Sachin