Hi, On Thu 22 Aug 19, 11:38, Ezequiel Garcia wrote: > On Thu, 2019-08-22 at 15:47 +0200, Hans Verkuil wrote: > > On 8/22/19 1:54 PM, Paul Kocialkowski wrote: > > > Hi, > > > > > > On Mon 19 Aug 19, 17:53, Hans Verkuil wrote: > > > > On 8/19/19 2:41 PM, Paul Kocialkowski wrote: > > > > > Hi, > > > > > > > > > > On Fri 16 Aug 19, 13:01, Ezequiel Garcia wrote: > > > > > > The V4L2_PIX_FMT_H264_SLICE_RAW name was originally suggested > > > > > > because the pixel format would represent H264 slices without any > > > > > > start code. > > > > > > > > > > > > However, as we will now introduce a start code menu control, > > > > > > give the pixel format a more meaningful name, while it's > > > > > > still early enough to do so. > > > > > > > > > > I definitely agree that SLICE_RAW is not the suffix we are looking for, but I'm > > > > > not sure that _SLICE is self-describing given that we can operate either > > > > > per-frame or per-slice, and _SLICE sort of implies the latter. Also, VP8 uses > > > > > _FRAME to clearly indicate that it operates per-frame. > > > > > > > > Well, VP8 doesn't support slices at all. > > > > > > > > > In addition, the _SLICE suffix is used by MPEG-2 in the stable API. Since we > > > > > > > > Regarding MPEG-2: while it has a concept of slices, it is my understanding > > > > that you never process slices separately, but only full pictures. I may be > > > > wrong here. > > > > > > I don't think that is the case since ffmpeg clearly implements decoding on a > > > per-slice basis (mpeg_decode_slice). > > > > > > Information is also passed on a per-slice basis to VAAPI > > > (vaapi_mpeg2_decode_slice) with a distinct data buffer and slice parameter > > > buffer for each slice. Among other things, it contains the vertical and > > > horizontal positions for the slice, which we can set in the hardware. > > > > > > > > certainly want MPEG-2 to allow per-slice and per-frame decoding as well as > > > > > H.264 and that the _SLICE format is specified to be the broken "concatenated > > > > > slices" that cedrus expects, we probably want to use another suffix. This way, > > > > > we could deprecated MPEG2_SLICE and introduce a new format for MPEG-2 that would > > > > > have consistent naming with the other mpeg formats. > > > > > > > > I actually think that H264_SLICE is a decent name. > > > > > > > > I'm less sure about MPEG2_SLICE since I am not sure if it means the same as > > > > a H264 slice. > > > > > > The main problem I see is that we have already specified MPEG2_SLICE in a way > > > that is incompatible with the future improvments we want to bring to the API: > > > " The output buffer must contain the appropriate number of macroblocks to > > > decode a full corresponding frame to the matching capture buffer." > > > > > > So I only see two possibilities: either we decide to change the specification > > > of the pixel format and we can keep using the _SLICE suffix, either we need to > > > introduce a new pixel format with another suffix, which should also be reflected > > > on other MPEG formats for consistency. Then we can deprecate MPEG2_SLICE and > > > have drivers stop using it. > > > > > > What do you think? > > > > I'd change the specification of the pixel format. So MPEG2_SLICE now supports > > multiple slices if the hardware supports it as well. > > > > We would need an MPEG2_DECODING_MODE control as well, that currently would > > read FRAME based only. > > > > That's exactly what I was about to suggest. Sounds perfect then! I have started looking at the start-code situation to see if it shares similarities with H.264, but did not go far enough to reach any conclusion on that aspect. Cheers, Paul -- Paul Kocialkowski, Bootlin Embedded Linux and kernel engineering https://bootlin.com
Attachment:
signature.asc
Description: PGP signature
