From: Paul B Mahol <onemda@gmail.com>
To: FFmpeg development discussions and patches <ffmpeg-devel@ffmpeg.org>
Subject: Re: [FFmpeg-devel] [PATCH] avfilter/src_movie: activate & dr
Date: Fri, 19 May 2023 10:07:20 +0200
Message-ID: <CAPYw7P7f1wAR6O4tQzJWiK=Af4_D0GavMBJzOD4iMNziDJ+OfA@mail.gmail.com> (raw)
In-Reply-To: <CAPYw7P6SZ0rH8f00RJ_6vGbZOP4sCX3M+G3JrxxRBGHPyvgGsQ@mail.gmail.com>
On Fri, May 19, 2023 at 9:52 AM Paul B Mahol <onemda@gmail.com> wrote:
>
>
> On Thu, May 18, 2023 at 8:01 PM James Almer <jamrial@gmail.com> wrote:
>
>> On 5/18/2023 2:44 PM, Paul B Mahol wrote:
>> > On Thu, May 18, 2023 at 7:31 PM James Almer <jamrial@gmail.com> wrote:
>> >
>> >> On 5/18/2023 7:22 AM, Paul B Mahol wrote:
>> >>> From af73b69a0be9033fddf222b6e9ac60799de85691 Mon Sep 17 00:00:00
>> 2001
>> >>> From: Paul B Mahol <onemda@gmail.com>
>> >>> Date: Mon, 15 May 2023 21:54:25 +0200
>> >>> Subject: [PATCH 26/27] avfilter/src_movie: dr support
>> >>>
>> >>> Signed-off-by: Paul B Mahol <onemda@gmail.com>
>> >>> ---
>> >>> libavfilter/src_movie.c | 44
>> +++++++++++++++++++++++++++++++++++++++--
>> >>> 1 file changed, 42 insertions(+), 2 deletions(-)
>> >>>
>> >>> diff --git a/libavfilter/src_movie.c b/libavfilter/src_movie.c
>> >>> index 5937613d13..a2ecc5a625 100644
>> >>> --- a/libavfilter/src_movie.c
>> >>> +++ b/libavfilter/src_movie.c
>> >>> @@ -23,7 +23,6 @@
>> >>> * @file
>> >>> * movie video source
>> >>> *
>> >>> - * @todo use direct rendering (no allocation of a new frame)
>> >>> * @todo support a PTS correction mechanism
>> >>> */
>> >>>
>> >>> @@ -156,6 +155,43 @@ static AVStream *find_stream(void *log,
>> >> AVFormatContext *avf, const char *spec)
>> >>> return found;
>> >>> }
>> >>>
>> >>> +static int get_buffer(AVCodecContext *avctx, AVFrame *frame, int
>> flags)
>> >>> +{
>> >>> + int linesize_align[AV_NUM_DATA_POINTERS];
>> >>> + AVFilterLink *outlink = frame->opaque;
>> >>> + int w, h, ow, oh;
>> >>> + AVFrame *new;
>> >>> +
>> >>> + h = oh = frame->height;
>> >>> + w = ow = frame->width;
>> >>> +
>> >>> + if (!(avctx->codec->capabilities & AV_CODEC_CAP_DR1))
>> >>> + return avcodec_default_get_buffer2(avctx, frame, flags);
>> >>> +
>> >>> + switch (avctx->codec_type) {
>> >>> + case AVMEDIA_TYPE_VIDEO:
>> >>> + avcodec_align_dimensions2(avctx, &w, &h, linesize_align);
>> >>> + new = ff_default_get_video_buffer(outlink, w, h);
>> >>> + break;
>> >>> + case AVMEDIA_TYPE_AUDIO:
>> >>> + new = ff_default_get_audio_buffer(outlink,
>> frame->nb_samples);
>> >>
>> >> I wonder, are these two functions thread safe?
>> >> AVCodecContext.get_buffer2() no longer supports non-thread safe
>> >> callbacks, and these seem to allocate and replace the pool inside
>> >> outlink as required. If called from more than one thread with different
>> >> frame parameters, wouldn't there be a race?
>> >>
>> >
>> > I encountered no issues in my testing.
>>
>> Try a sample that changes dimensions every other frame, like
>> vp9-test-vectors/vp90-2-05-resize.ivf, and make sure to set filter
>> threads so it's propagated to the src_movie internal decoder, to some
>> value like 4 or higher. Run it under tsan.
>>
>>
> I tried that, and also own .h264 file hash decoding is always same.
> tsan shows warnings for both without and with this filter and frame
> threading.
>
Only issue I found is if you use tmix filter after: ffmpeg -f lavfi -i
movie=resolution_change_name.file,tmix
It will produce invalid output with MD5 hash different each time, same as
with threads=1
But exact same issue is without this 2 patches, and present in current
version of FFmpeg.
Nice workaround is just inserting scale filter between movie and tmix.
>
>> >
>> >
>> >>
>> >>> + break;
>> >>> + default:
>> >>> + return -1;
>> >>> + }
>> >>> +
>> >>> + av_frame_copy_props(new, frame);
>> >>> + av_frame_unref(frame);
>> >>> + av_frame_move_ref(frame, new);
>> >>> + av_frame_free(&new);
>> >>> +
>> >>> + frame->opaque = outlink;
>> >>> + frame->width = ow;
>> >>> + frame->height = oh;
>> >>> +
>> >>> + return 0;
>> >>> +}
>> >>> +
>> >>> static int open_stream(AVFilterContext *ctx, MovieStream *st, int
>> >> dec_threads)
>> >>> {
>> >>> const AVCodec *codec;
>> >>> @@ -171,6 +207,8 @@ static int open_stream(AVFilterContext *ctx,
>> >> MovieStream *st, int dec_threads)
>> >>> if (!st->codec_ctx)
>> >>> return AVERROR(ENOMEM);
>> >>>
>> >>> + st->codec_ctx->flags |= AV_CODEC_FLAG_COPY_OPAQUE;
>> >>> + st->codec_ctx->get_buffer2 = get_buffer;
>> >>> ret = avcodec_parameters_to_context(st->codec_ctx,
>> >> st->st->codecpar);
>> >>> if (ret < 0)
>> >>> return ret;
>> >>> @@ -479,8 +517,10 @@ static int movie_decode_packet(AVFilterContext
>> *ctx)
>> >>> /* send the packet to its decoder, if any */
>> >>> pkt_out_id = pkt.stream_index > movie->max_stream_index ? -1 :
>> >>> movie->out_index[pkt.stream_index];
>> >>> - if (pkt_out_id >= 0)
>> >>> + if (pkt_out_id >= 0) {
>> >>> + pkt.opaque = ctx->outputs[pkt_out_id];
>> >>> ret = avcodec_send_packet(movie->st[pkt_out_id].codec_ctx,
>> >> &pkt);
>> >>> + }
>> >>> av_packet_unref(&pkt);
>> >>>
>> >>> return ret;
>> >>> --
>> >>> 2.39.1
>> >>>
>> >> _______________________________________________
>> >> ffmpeg-devel mailing list
>> >> ffmpeg-devel@ffmpeg.org
>> >> https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
>> >>
>> >> To unsubscribe, visit link above, or email
>> >> ffmpeg-devel-request@ffmpeg.org with subject "unsubscribe".
>> >>
>> > _______________________________________________
>> > ffmpeg-devel mailing list
>> > ffmpeg-devel@ffmpeg.org
>> > https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
>> >
>> > To unsubscribe, visit link above, or email
>> > ffmpeg-devel-request@ffmpeg.org with subject "unsubscribe".
>> _______________________________________________
>> ffmpeg-devel mailing list
>> ffmpeg-devel@ffmpeg.org
>> https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
>>
>> To unsubscribe, visit link above, or email
>> ffmpeg-devel-request@ffmpeg.org with subject "unsubscribe".
>>
>
_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
To unsubscribe, visit link above, or email
ffmpeg-devel-request@ffmpeg.org with subject "unsubscribe".
next prev parent reply other threads:[~2023-05-19 8:08 UTC|newest]
Thread overview: 9+ messages / expand[flat|nested] mbox.gz Atom feed top
2023-05-18 10:22 Paul B Mahol
2023-05-18 17:31 ` James Almer
2023-05-18 17:44 ` Paul B Mahol
2023-05-18 18:01 ` James Almer
2023-05-19 7:52 ` Paul B Mahol
2023-05-19 8:07 ` Paul B Mahol [this message]
2023-05-22 14:48 ` Anton Khirnov
2023-05-22 14:55 ` James Almer
2023-05-22 15:00 ` James Almer
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to='CAPYw7P7f1wAR6O4tQzJWiK=Af4_D0GavMBJzOD4iMNziDJ+OfA@mail.gmail.com' \
--to=onemda@gmail.com \
--cc=ffmpeg-devel@ffmpeg.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Git Inbox Mirror of the ffmpeg-devel mailing list - see https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
This inbox may be cloned and mirrored by anyone:
git clone --mirror https://master.gitmailbox.com/ffmpegdev/0 ffmpegdev/git/0.git
# If you have public-inbox 1.1+ installed, you may
# initialize and index your mirror using the following commands:
public-inbox-init -V2 ffmpegdev ffmpegdev/ https://master.gitmailbox.com/ffmpegdev \
ffmpegdev@gitmailbox.com
public-inbox-index ffmpegdev
Example config snippet for mirrors.
AGPL code for this site: git clone https://public-inbox.org/public-inbox.git