Dynamic array control support: please test!

Dynamic array control support: please test!

[Hans Verkuil]

Hi Ezequiel, John,

After creating extensive new compliance tests for this feature I am now
confident enough about the implementation.

You can find it here:

https://git.linuxtv.org/hverkuil/media_tree.git/log/?h=ctrl-refactor

There are two items that I want to fix before I will post this:

1) The new flag needs to be documented
2) I think there are some simplifications possible w.r.t. storing the
   size of the new array, I want to look at that a bit more.

In the meantime it would be great if some testing of this series can be
done with real drivers instead of just vivid.

Regards,

	Hans

Re: Dynamic array control support: please test!

[John Cox]

Hi

>Hi Ezequiel, John,
>
>After creating extensive new compliance tests for this feature I am now
>confident enough about the implementation.
>
>You can find it here:
>
>https://git.linuxtv.org/hverkuil/media_tree.git/log/?h=ctrl-refactor
>
>There are two items that I want to fix before I will post this:
>
>1) The new flag needs to be documented
>2) I think there are some simplifications possible w.r.t. storing the
>   size of the new array, I want to look at that a bit more.
>
>In the meantime it would be great if some testing of this series can be
>done with real drivers instead of just vivid.

Well I finally managed to put together your patch (in 5.10), tweaked
driver & ffmpeg to have a variable noof slice headers and it all seems
to work well. With the code I have it doesn't give a significant
improvement in performance over 1 slice at a time but that is probably
because I have multithreaded userland code and take liberties with
buffer returns from the driver that Ezequiel disaproves of (but my
ffmpeg code is happy with).

It definitely saves the driver having to alloc any intermediate buffers
to stash the bitstream in.

Many thanks

John Cox

Re: Dynamic array control support: please test!

[Ezequiel Garcia]

Hi John,
 
On Thursday, April 29, 2021 14:56 -03, John Cox <jc@kynesim.co.uk> wrote: 
 
> Hi
> 
> >Hi Ezequiel, John,
> >
> >After creating extensive new compliance tests for this feature I am now
> >confident enough about the implementation.
> >
> >You can find it here:
> >
> >https://git.linuxtv.org/hverkuil/media_tree.git/log/?h=ctrl-refactor
> >
> >There are two items that I want to fix before I will post this:
> >
> >1) The new flag needs to be documented
> >2) I think there are some simplifications possible w.r.t. storing the
> >   size of the new array, I want to look at that a bit more.
> >
> >In the meantime it would be great if some testing of this series can be
> >done with real drivers instead of just vivid.
> 
> Well I finally managed to put together your patch (in 5.10), tweaked
> driver & ffmpeg to have a variable noof slice headers and it all seems
> to work well. With the code I have it doesn't give a significant
> improvement in performance over 1 slice at a time but that is probably
> because I have multithreaded userland code and take liberties with
> buffer returns from the driver that Ezequiel disaproves of (but my
> ffmpeg code is happy with).
> 

I don't think I have ever disapproved anything (and if I made it sound like that, it was a mistake). Quite the opposite, I am more than supportive to see get more drivers merged upstream.

Thanks!
Ezequiel

Re: Dynamic array control support: please test!

[John Cox]

Hi Ezequiel

>Hi John,
> 
>On Thursday, April 29, 2021 14:56 -03, John Cox <jc@kynesim.co.uk> wrote: 
> 
>> Hi
>> 
>> >Hi Ezequiel, John,
>> >
>> >After creating extensive new compliance tests for this feature I am now
>> >confident enough about the implementation.
>> >
>> >You can find it here:
>> >
>> >https://git.linuxtv.org/hverkuil/media_tree.git/log/?h=ctrl-refactor
>> >
>> >There are two items that I want to fix before I will post this:
>> >
>> >1) The new flag needs to be documented
>> >2) I think there are some simplifications possible w.r.t. storing the
>> >   size of the new array, I want to look at that a bit more.
>> >
>> >In the meantime it would be great if some testing of this series can be
>> >done with real drivers instead of just vivid.
>> 
>> Well I finally managed to put together your patch (in 5.10), tweaked
>> driver & ffmpeg to have a variable noof slice headers and it all seems
>> to work well. With the code I have it doesn't give a significant
>> improvement in performance over 1 slice at a time but that is probably
>> because I have multithreaded userland code and take liberties with
>> buffer returns from the driver that Ezequiel disaproves of (but my
>> ffmpeg code is happy with).
>> 
>
>I don't think I have ever disapproved anything (and if I made it sound like that, it was a mistake). Quite the opposite, I am more than supportive to see get more drivers merged upstream.

Yes - sorry - my mistake - I was thinking of one person and typed
another. And in any case, even if I had the right name, that wasn't a
constructive remark on my part.

Please accept my apologies

John Cox

>Thanks!
>Ezequiel

Re: Dynamic array control support: please test!

[John Cox]

Hi

>Hi Ezequiel, John,
>
>After creating extensive new compliance tests for this feature I am now
>confident enough about the implementation.
>
>You can find it here:
>
>https://git.linuxtv.org/hverkuil/media_tree.git/log/?h=ctrl-refactor
>
>There are two items that I want to fix before I will post this:
>
>1) The new flag needs to be documented
>2) I think there are some simplifications possible w.r.t. storing the
>   size of the new array, I want to look at that a bit more.
>
>In the meantime it would be great if some testing of this series can be
>done with real drivers instead of just vivid.

Have dynamic arrays got lost or have they made it in under another name?
The ctrl split made it, but dynamic arrays less obviously.

Regards

John Cox