Re: [PATCH 00/33] Qualcomm video decoder/encoder driver

[Dmitry Baryshkov]

Hi Stan,

On Mon, 14 Aug 2023 at 15:58, Stanimir Varbanov
<stanimir.k.varbanov@xxxxxxxxx> wrote:
>
> Hi Dmitry,
>
> On 28.07.23 г. 17:01 ч., Dmitry Baryshkov wrote:
> > On 28/07/2023 16:23, Vikash Garodia wrote:
> >> This patch series introduces support for Qualcomm new video acceleration
> >> hardware architecture, used for video stream decoding/encoding. This
> >> driver
> >> is based on new communication protocol between video hardware and
> >> application
> >> processor.
> >>
> >> This driver comes with below capabilities:
> >> - V4L2 complaint video driver with M2M and STREAMING capability.
> >> - Supports H264, H265, VP9 decoders.
> >> - Supports H264, H265 encoders.
> >
> > Please describe, why is it impossible to support this hardware in the
> > venus driver. We do not usually add new drivers for the new generations
> > of the hardware, unless it is fully incompatible with the previous
> > generations. Let me point you to camss or drm/msm drivers. They have
> > successfully solved the issue of supporting multiple generations of the
> > hardware in the same driver.
> >
> > Unless the "iris3" is completely different from all the previous
> > generations, I strongly suggest spending time on restructuring existing
> > venus driver and then adding support for the new hardware there instead
> > of dumping out something completely new.
>
> AFAIK the major differences are HW IP and firmware interface (by
> firmware interface I mean a protocol, API and API behavior). The
> firmware and its interface has been re-written to align closely with the
> current v4l2 specs for encoders/decoders state machines [1][2]. On the
> other side current mainline Venus driver firmware is following interface
> similar to OpenMAX.
>
> There are incompatibilities between both firmware interfaces which
> cannot easily combined in a common driver. Even if there is a
> possibility to do that it will lead us to a unreadable driver source
> code and maintenance burden.

Thank you for your explanation!

If the hardware is more or less the same, then the existing venus
driver should be refactored and split into hardware driver and the
firmware interface. Then iris3 can come up as a second driver
implementing support for new firmware interface but utilising common
hardware-related code.

> Vikash, could elaborate more on firmware interface differences.

Do we have any details on firmware versions that implement older
(OpenMAX-like) interface vs versions implementing new (v4l2-like)
interface?

> [1]
> https://www.kernel.org/doc/html/latest/userspace-api/media/v4l/dev-decoder.html
>
> [2]
> https://www.kernel.org/doc/html/latest/userspace-api/media/v4l/dev-encoder.html

-- 
With best wishes
Dmitry