[libcamera-devel] [PATCH v3 1/4] ipa: ipu3: Separate out frame context queue as a distinct class

[Umang Jain]

Hi Jacopo

On 6/20/22 18:44, Jacopo Mondi wrote:
> Hi Umang
>
> On Mon, Jun 20, 2022 at 06:35:07PM +0530, Umang Jain wrote:
>> Hi Jacopo
>>
>> On 6/20/22 15:14, Jacopo Mondi wrote:
>>> Hi Umang,
>>>
>>> On Mon, Jun 20, 2022 at 02:10:21AM +0530, Umang Jain via libcamera-devel wrote:
>>>> +
>>>> +/**
>>>> + * \brief Sets the size of the FCQueue
>>>> + * \param[in] size The intended size of the FCQueue
>>>> + */
>>>> +void FCQueue::setSize(uint32_t size)
>>> Will the size be changed at run time or is it fixed ? Because if
>>> that's the case this could be made a compile-time parameter..
>>
>> FCQueue should re-configuration at run time, in the sense, it can be
>> re-configured to change it's size between start() > stop() > configure() >
>> stop() cycles.
>>
>> That's what I have tried to address.
>>
> Could you be patient and explain me what is the use case for changing the
> FCQueue size (ie the max number of requests in flight, which to me is
> a camera static property).


Maybe not, I might be wrong yes.

But it's closely tied to the "look-ahead" frames of the algorithms which 
makes me think away from Camera properties.

The "look-ahead" frames for algorithms might be closely related to 
maximum delay of the camera sensor controls so there it seems more 
closer to camera property (were you inferring that from that start?)

I haven't really given a thought, but we can always fix it (to a compile 
time constant) if we want to. I'll wait for more feedback though