virtio-video device IDs

virtio-video device IDs

[Alexander Gordeev]

Hi,

Currently there are two device IDs assigned for the future virtio video device:
 * 30 for the encoder
 * 31 for the decoder
I received feedback from our internal reviewers that in some cases it would be better to use as few virtqueues as possible. So if a single piece of HW supports both encoder and decoder functions, then it is better to have a single virtual device, that provides both functions as well. The specific function can be selected per stream then.
Is it OK to make changes to the reserved IDs?
Are there any other concerns?

Thanks

RE: virtio-video device IDs

[Parav Pandit]

> From: Alexander Gordeev <quic_agordeev@quicinc.com>
> Sent: Friday, December 20, 2024 10:17 PM
> 
> Hi,
> 
> Currently there are two device IDs assigned for the future virtio video device:
>  * 30 for the encoder
>  * 31 for the decoder
> I received feedback from our internal reviewers that in some cases it would
> be better to use as few virtqueues as possible. So if a single piece of HW
> supports both encoder and decoder functions, then it is better to have a
> single virtual device, that provides both functions as well. The specific
> function can be selected per stream then.
> Is it OK to make changes to the reserved IDs?
> Are there any other concerns?
> 
> Thanks

For the new device usually, there is a new device id.
May be you can explore how to merge the functionality using capabilities or feature bits to 
a. either extend encoder to have additional decoder
or
b. extend decoded to have the encoder

and than the description can be changed.

Or 
c. if the new device has major functionality and performance, may be its own.

I believe the details would be useful to find the suitable option from the 3 choices.

Re: virtio-video device IDs

[Alexander Gordeev]

On 22/12/2024 15:14, Parav Pandit wrote:
>> From: Alexander Gordeev <quic_agordeev@quicinc.com>
>> Sent: Friday, December 20, 2024 10:17 PM
>>
>> Hi,
>>
>> Currently there are two device IDs assigned for the future virtio video device:
>>  * 30 for the encoder
>>  * 31 for the decoder
>> I received feedback from our internal reviewers that in some cases it would
>> be better to use as few virtqueues as possible. So if a single piece of HW
>> supports both encoder and decoder functions, then it is better to have a
>> single virtual device, that provides both functions as well. The specific
>> function can be selected per stream then.
>> Is it OK to make changes to the reserved IDs?
>> Are there any other concerns?
>>
>> Thanks
> 
> For the new device usually, there is a new device id.
> May be you can explore how to merge the functionality using capabilities or feature bits to 
> a. either extend encoder to have additional decoder
> or
> b. extend decoded to have the encoder
> 
> and than the description can be changed.
> 
> Or 
> c. if the new device has major functionality and performance, may be its own.
> 
> I believe the details would be useful to find the suitable option from the 3 choices.

I don't understand option 3. Perhaps there are typos.
At the moment I think we'll choose option 1. Thanks!

Re: virtio-video device IDs

[Michael S. Tsirkin]

On Fri, Dec 20, 2024 at 05:47:11PM +0100, Alexander Gordeev wrote:
> Hi,
> 
> Currently there are two device IDs assigned for the future virtio video device:
>  * 30 for the encoder
>  * 31 for the decoder
> I received feedback from our internal reviewers that in some cases it would be better to use as few virtqueues as possible. So if a single piece of HW supports both encoder and decoder functions, then it is better to have a single virtual device, that provides both functions as well. The specific function can be selected per stream then.
> Is it OK to make changes to the reserved IDs?
> Are there any other concerns?
> 
> Thanks

Basically, we need to make sure there are no conflicts.
The IDs were reserved by google and chromium guys.
If you want to make changes, get in touch with them
and get their ACK please.

But generally, it's easier to just reserve another ID
for an encoder/decoder device.
We are not short on these.

-- 
MST

Re: virtio-video device IDs

[Alexander Gordeev]

On 23/12/2024 11:57, Michael S. Tsirkin wrote:
> On Fri, Dec 20, 2024 at 05:47:11PM +0100, Alexander Gordeev wrote:
>> Hi,
>>
>> Currently there are two device IDs assigned for the future virtio video device:
>>  * 30 for the encoder
>>  * 31 for the decoder
>> I received feedback from our internal reviewers that in some cases it would be better to use as few virtqueues as possible. So if a single piece of HW supports both encoder and decoder functions, then it is better to have a single virtual device, that provides both functions as well. The specific function can be selected per stream then.
>> Is it OK to make changes to the reserved IDs?
>> Are there any other concerns?
>>
>> Thanks
> 
> Basically, we need to make sure there are no conflicts.
> The IDs were reserved by google and chromium guys.
> If you want to make changes, get in touch with them
> and get their ACK please.
> 
> But generally, it's easier to just reserve another ID
> for an encoder/decoder device.
> We are not short on these.
> 

Thanks! Yes, I think we'll just reserve another ID indeed.
Maybe the old ones would become useful again later.