On Tue, Mar 06, 2018 at 01:15:55PM -0800, Aman Gupta wrote: > From: Aman Gupta <[email protected]> > > The default behavior of the mediacodec decoder before this commit > was to delay flushes until all pending hardware frames were > returned to the decoder. This was useful for certain types of > applications, but was unexpected behavior for others. > > The new default behavior with this commit is now to execute > flushes immediately to invalidate all pending frames. The old > behavior can be enabled by setting delay_flush=1. > > With the new behavior, video players implementing seek can simply > call flush on the decoder without having to worry about whether > they have one or more mediacodec frames still buffered in their > rendering pipeline. Previously, all these frames had to be > explictly freed (or rendered) before the seek/flush would execute. > > The new behavior matches the behavior of all other lavc decoders, > reducing the amount of special casing required when using the > mediacodec decoder. > --- > libavcodec/mediacodec.c | 2 +- > libavcodec/mediacodecdec.c | 23 +++++++++++++++++++++++ > libavcodec/mediacodecdec_common.c | 11 ++++++++--- > libavcodec/mediacodecdec_common.h | 4 ++++ > 4 files changed, 36 insertions(+), 4 deletions(-) > > diff --git a/libavcodec/mediacodec.c b/libavcodec/mediacodec.c > index d492eefe0b..bf1b7477f1 100644 > --- a/libavcodec/mediacodec.c > +++ b/libavcodec/mediacodec.c > @@ -91,7 +91,7 @@ int av_mediacodec_release_buffer(AVMediaCodecBuffer > *buffer, int render) > MediaCodecDecContext *ctx = buffer->ctx; > int released = atomic_fetch_add(&buffer->released, 1); > > - if (!released) { > + if (!released && (ctx->delay_flush || buffer->serial == > atomic_load(&ctx->serial))) { > return ff_AMediaCodec_releaseOutputBuffer(ctx->codec, buffer->index, > render); > } > > diff --git a/libavcodec/mediacodecdec.c b/libavcodec/mediacodecdec.c > index 0fe14846c3..3582a1377d 100644 > --- a/libavcodec/mediacodecdec.c > +++ b/libavcodec/mediacodecdec.c > @@ -41,10 +41,14 @@ > > typedef struct MediaCodecH264DecContext { > > + AVClass *avclass; > + > MediaCodecDecContext *ctx; > > AVPacket buffered_pkt; > > + int delay_flush; > + > } MediaCodecH264DecContext; > > static av_cold int mediacodec_decode_close(AVCodecContext *avctx) > @@ -366,6 +370,8 @@ static av_cold int mediacodec_decode_init(AVCodecContext > *avctx) > goto done; > } > > + s->ctx->delay_flush = s->delay_flush; > + > if ((ret = ff_mediacodec_dec_init(avctx, s->ctx, codec_mime, format)) < > 0) { > s->ctx = NULL; > goto done; > @@ -485,7 +491,24 @@ static const AVCodecHWConfigInternal > *mediacodec_hw_configs[] = { > NULL > }; > > +#define OFFSET(x) offsetof(MediaCodecH264DecContext, x) > +#define VD AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_DECODING_PARAM > +static const AVOption ff_mediacodec_vdec_options[] = { > + { "delay_flush", "Delay flush until hw output buffers are returned to > the decoder", > + OFFSET(delay_flush), AV_OPT_TYPE_BOOL, {.i64 = 0}, 0, > 1, VD }, > + { NULL } > +}; > + > +#define DECLARE_MEDIACODEC_VCLASS(short_name) \ > +static const AVClass ff_##short_name##_mediacodec_dec_class = { \ > + .class_name = #short_name "_mediacodec", \ > + .item_name = av_default_item_name, \ > + .option = ff_mediacodec_vdec_options, \ > + .version = LIBAVUTIL_VERSION_INT, \ > +}; > + > #define DECLARE_MEDIACODEC_VDEC(short_name, full_name, codec_id, bsf) > \ > +DECLARE_MEDIACODEC_VCLASS(short_name) > \ > AVCodec ff_##short_name##_mediacodec_decoder = { > \ > .name = #short_name "_mediacodec", > \ > .long_name = NULL_IF_CONFIG_SMALL(full_name " Android MediaCodec > decoder"), \
The priv_class assignment is missing: + .priv_class = &ff_##short_name##_mediacodec_dec_class, \ You will also need some kind of bump in the libavcodec library as well as an APIChanges entry. Other than that the patch looks good to me. Thanks, [...] -- Matthieu B. _______________________________________________ ffmpeg-devel mailing list [email protected] http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
