On Thu, Jun 27, 2019 at 08:59:12PM +0800, Zhong Li wrote:
> It will be reused in the following mjpeg_parser patch
>
> Signed-off-by: Zhong Li <zhong...@intel.com>
> ---
> Mark Thompson: This seems suspicious - MJPEG is generally 4:2:2 (e.g. UVC
> requires it), so I would expect a 4:2:2 format to be the default here? (Or
> maybe a longer list - VAAPI certainly supports 4:2:2, 4:2:0 and 4:4:4 on the
> same hardware.)
> Zhong: libmfx can support jpeg baseline profile with more output formats, but
> current ffmpeg-qsv decoder/vpp can't. Will extend supported format list as
> separated patch.
>
> libavcodec/mjpegdec.c | 37 ++++++++++++++++++++++++++++---------
> libavcodec/mjpegdec.h | 4 ++++
> 2 files changed, 32 insertions(+), 9 deletions(-)
>
> diff --git a/libavcodec/mjpegdec.c b/libavcodec/mjpegdec.c
> index a65bc8d..5da66bb 100644
> --- a/libavcodec/mjpegdec.c
> +++ b/libavcodec/mjpegdec.c
> @@ -157,6 +157,8 @@ av_cold int ff_mjpeg_decode_init(AVCodecContext *avctx)
> s->start_code = -1;
> s->first_picture = 1;
> s->got_picture = 0;
> + s->reinit_idct = 0;
> + s->size_change = 0;
> s->org_height = avctx->coded_height;
> avctx->chroma_sample_location = AVCHROMA_LOC_CENTER;
> avctx->colorspace = AVCOL_SPC_BT470BG;
> @@ -302,9 +304,9 @@ int ff_mjpeg_decode_dht(MJpegDecodeContext *s)
> return 0;
> }
>
> -int ff_mjpeg_decode_sof(MJpegDecodeContext *s)
> +int ff_mjpeg_decode_header(MJpegDecodeContext *s)
> {
> - int len, nb_components, i, width, height, bits, ret, size_change;
> + int len, nb_components, i, width, height, bits, ret;
> unsigned pix_fmt_id;
> int h_count[MAX_COMPONENTS] = { 0 };
> int v_count[MAX_COMPONENTS] = { 0 };
> @@ -324,7 +326,7 @@ int ff_mjpeg_decode_sof(MJpegDecodeContext *s)
> if (s->avctx->bits_per_raw_sample != bits) {
> av_log(s->avctx, s->avctx->bits_per_raw_sample > 0 ? AV_LOG_INFO :
> AV_LOG_DEBUG, "Changing bps from %d to %d\n", s->avctx->bits_per_raw_sample,
> bits);
> s->avctx->bits_per_raw_sample = bits;
> - init_idct(s->avctx);
> + s->reinit_idct = 1;
> }I think the avctx->bits_per_raw_sample value should stay in sync with the initialized idct This patch would allow the bits_per_raw_sample to change and then a subsequent error to skip init_idct() maybe this is ok as it would be probably called later in a subsequent frame but i think its better if they stay closer together or there is nothing between them that can cause ine to exeucute without the other [...] -- Michael GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB Whats the most studid thing your enemy could do ? Blow himself up Whats the most studid thing you could do ? Give up your rights and freedom because your enemy blew himself up.
signature.asc
Description: PGP signature
_______________________________________________ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org https://ffmpeg.org/mailman/listinfo/ffmpeg-devel To unsubscribe, visit link above, or email ffmpeg-devel-requ...@ffmpeg.org with subject "unsubscribe".
