Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Message-ID: <c6a222be5eee962581cf5dcb9a1473cf45ff303c.camel@collabora.com>
Subject: Re: Stateless Encoding uAPI Discussion and Proposal
From:   Nicolas Dufresne <nicolas.dufresne@collabora.com>
To:     Michael Grzeschik <mgr@pengutronix.de>,
        Paul Kocialkowski <paul.kocialkowski@bootlin.com>
Cc:     linux-kernel@vger.kernel.org, linux-media@vger.kernel.org,
        Hans Verkuil <hverkuil@xs4all.nl>,
        Sakari Ailus <sakari.ailus@iki.fi>,
        Andrzej Pietrasiewicz <andrzej.p@collabora.com>,
        Michael Tretter <m.tretter@pengutronix.de>,
        Jernej =?UTF-8?Q?=C5=A0krabec?= <jernej.skrabec@gmail.com>,
        Chen-Yu Tsai <wens@csie.org>,
        Samuel Holland <samuel@sholland.org>,
        Thomas Petazzoni <thomas.petazzoni@bootlin.com>
Date:   Mon, 24 Jul 2023 10:03:12 -0400
In-Reply-To: <20230721181951.GL12001@pengutronix.de>
References: <ZK2NiQd1KnraAr20@aptenodytes>
         <20230721181951.GL12001@pengutronix.de>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
User-Agent: Evolution 3.48.4 (3.48.4-1.fc38) 
MIME-Version: 1.0
Precedence: bulk

Le vendredi 21 juillet 2023 =C3=A0 20:19 +0200, Michael Grzeschik a =C3=A9c=
rit=C2=A0:
> > As a result, we cannot expect that any given encoder is able to produce=
 frames
> > for any set of headers. Reporting related constraints and limitations (=
beyond
> > profile/level) seems quite difficult and error-prone.
> >=20
> > So it seems that keeping header generation in-kernel only (close to whe=
re the
> > hardware is actually configured) is the safest approach.
>=20
> For the case with the rkvenc, the headers are also not created by the
> kernel driver. Instead we use the gst_h264_bit_writer_sps/pps functions
> that are part of the codecparsers module.

One level of granularity we can add is split headers (like SPS/PPS) and
slice/frame headers. It remains that in some cases, like HEVC, when the sli=
ce
header is byte aligned, it can be nice to be able to handle it at applicati=
on
side in order to avoid limiting SVC support (and other creative features) b=
y our
API/abstraction limitations. I think a certain level of "per CODEC" reasoni=
ng is
also needed. Just like, I would not want to have to ask the kernel to gener=
ate
user data SEI and other in-band data.

regards,
Nicolas