[FFmpeg-devel] Re: [FFmpeg-cvslog] [ffmpeg] branch master updated. 92f2f9ea5c avcodec/d3d12va_encode: D3D12 AV1 encoding support

[FFmpeg-devel] Re: [FFmpeg-cvslog] [ffmpeg] branch master updated. 92f2f9ea5c avcodec/d3d12va_encode: D3D12 AV1 encoding support

[Michael Niedermayer via ffmpeg-devel]

[-- Attachment #1.1: Type: text/plain, Size: 3971 bytes --]

Hi

On Wed, Nov 26, 2025 at 11:59:09AM +0200, ffmpeg-git--- via ffmpeg-cvslog wrote:
> The branch, master has been updated
>        via  92f2f9ea5c49e94814693861a7e9e47c993fca2e (commit)
>       from  81362b319ea7244d8d17110adfa59f10c7e78268 (commit)
> 
> 
> - Log -----------------------------------------------------------------
> commit 92f2f9ea5c49e94814693861a7e9e47c993fca2e
> Author:     Araz Iusubov <Primeadvice@gmail.com>
> AuthorDate: Mon Nov 10 17:23:25 2025 +0100
> Commit:     Tong Wu <wutong1208@outlook.com>
> CommitDate: Wed Nov 26 09:58:44 2025 +0000
> 
>     avcodec/d3d12va_encode: D3D12 AV1 encoding support
>     
>     Implement AV1 hardware encoding
>     using Direct3D 12 Video API (D3D12VA).
> 
[...]
>  int ff_d3d12va_encode_receive_packet(AVCodecContext *avctx, AVPacket *pkt);
> diff --git a/libavcodec/d3d12va_encode_av1.c b/libavcodec/d3d12va_encode_av1.c
> new file mode 100644
> index 0000000000..e7a115a2ee
> --- /dev/null
> +++ b/libavcodec/d3d12va_encode_av1.c
[...]
> +static int d3d12va_encode_av1_get_coded_data(AVCodecContext *avctx,
> +                                             D3D12VAEncodePicture *pic, AVPacket *pkt)
> +{
> +    int                   err = 0;
> +    uint8_t              *ptr = NULL;
> +    uint8_t      *mapped_data = NULL;
> +    size_t         total_size = 0;
> +    HRESULT                hr = S_OK;
> +    size_t    av1_pic_hd_size = 0;
> +    int tile_group_extra_size = 0;
> +    size_t            bit_len = 0;
> +    D3D12VAEncodeContext *ctx = avctx->priv_data;
> +
> +    char pic_hd_data[MAX_PARAM_BUFFER_SIZE] = { 0 };
> +
> +    err = d3d12va_encode_av1_get_buffer_size(avctx, pic, &total_size);
> +    if (err < 0)
> +        goto end;
> +
> +    // Update the picture header and calculate the picture header size
> +    memset(pic_hd_data, 0, sizeof(pic_hd_data));
> +    err = d3d12va_encode_av1_write_picture_header(avctx, pic, pic_hd_data, &av1_pic_hd_size);
> +    if (err < 0) {
> +        av_log(avctx, AV_LOG_ERROR, "Failed to write picture header: %d.\n", err);
> +        return err;
> +    }
> +    av1_pic_hd_size /= 8;
> +    av_log(avctx, AV_LOG_DEBUG, "AV1 picture header size: %zu bytes.\n", av1_pic_hd_size);
> +
> +
> +    tile_group_extra_size = (av_log2(total_size) + 7) / 7 + 1; // 1 byte for obu header, rest for tile group LEB128 size
> +    av_log(avctx, AV_LOG_DEBUG, "Tile group extra size: %d bytes.\n", tile_group_extra_size);
> +
> +    total_size += (pic->header_size + tile_group_extra_size + av1_pic_hd_size);
> +    av_log(avctx, AV_LOG_DEBUG, "Output buffer size %"SIZE_SPECIFIER"\n", total_size);
> +
> +    hr = ID3D12Resource_Map(pic->output_buffer, 0, NULL, (void **)&mapped_data);
> +    if (FAILED(hr)) {
> +        err = AVERROR_UNKNOWN;
> +        goto end;
> +    }
> +
> +    err = ff_get_encode_buffer(avctx, pkt, total_size, 0);
> +    if (err < 0)
> +        goto end;
> +    ptr = pkt->data;
> +
> +    memcpy(ptr, mapped_data, pic->header_size);
> +
> +    ptr += pic->header_size;
> +    mapped_data += pic->aligned_header_size;
> +    total_size -= pic->header_size;
> +
> +    memcpy(ptr, pic_hd_data, av1_pic_hd_size);
> +    ptr += av1_pic_hd_size;
> +    total_size -= av1_pic_hd_size;
> +    av_log(avctx, AV_LOG_DEBUG, "AV1 total_size after write picture header: %d.\n", total_size);
> +
> +    total_size -= tile_group_extra_size;
> +    err = d3d12va_encode_av1_write_tile_group(avctx, mapped_data, total_size, ptr, &bit_len);
> +    if (err < 0) {
> +        av_log(avctx, AV_LOG_ERROR, "Failed to write tile group: %d.\n", err);
> +        goto end;
> +    }

> +    assert((total_size + tile_group_extra_size) * 8 == bit_len);

this possibly should be av_assert*

thx

[...]
-- 
Michael     GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB

Observe your enemies, for they first find out your faults. -- Antisthenes

[-- Attachment #1.2: signature.asc --]
[-- Type: application/pgp-signature, Size: 195 bytes --]

[-- Attachment #2: Type: text/plain, Size: 163 bytes --]

_______________________________________________
ffmpeg-devel mailing list -- ffmpeg-devel@ffmpeg.org
To unsubscribe send an email to ffmpeg-devel-leave@ffmpeg.org

[FFmpeg-devel] Re: [FFmpeg-cvslog] [ffmpeg] branch master updated. 3d96d83a0a avformat/rawdec: set framerate in codec parameters

[Michael Niedermayer via ffmpeg-devel]

[-- Attachment #1.1: Type: text/plain, Size: 8345 bytes --]

Hi Gavin

this patch breaks decoding a video with a single image:

this works and creates our single image video
./ffmpeg -i lena.pnm /tmp/file.m2v

this fails after the patch:
./ffmpeg -i /tmp/file.m2v /tmp/file.jpg -y

[image2 @ 0x60dca024f140] Cannot write more than one file with the same name. Are you missing the -update option or a sequence pattern?
[vost#0:0/mjpeg @ 0x60dca024f3c0] Error submitting a packet to the muxer: Invalid argument
    Last message repeated 1 times
[out#0/image2 @ 0x60dca024f040] Error muxing a packet
[out#0/image2 @ 0x60dca024f040] Task finished with error code: -22 (Invalid argument)
[out#0/image2 @ 0x60dca024f040] Terminating thread with return code -22 (Invalid argument)
[out#0/image2 @ 0x60dca024f040] video:375KiB audio:0KiB subtitle:0KiB other streams:0KiB global headers:0KiB muxing overhead: unknown
frame=    2 fps=0.0 q=5.0 Lsize=N/A time=00:00:00.04 bitrate=N/A dup=1 drop=0 speed=0.455x elapsed=0:00:00.08
Conversion failed!

This also breaks:
./ffmpeg -i https://trac.ffmpeg.org/raw-attachment/ticket/3923/frame16.h264 test.jpg

rawdec has a user parameter called framerate which is set to 25 by default.
If your input is not 25, obviously then it should not be set to 25.

Your change does not achive a fallback of 25 if it cannot be determined,
as clearly shown in the example above where our single frame is turned into
2 frames, clearly showing there is a new inconsistancy.

I suggest to revert this patch (to avoid creating more bugs by a rushed attempt to
fix this) and attempt to fix your file by setting the user framerate parameter correctly

Independant of this, making it autodetect would be cool, yes, i agree
but i think we need more time, testing and no regressions

there where also suspicous changes in fate below:

On Fri, Nov 28, 2025 at 02:55:58AM +0200, ffmpeg-git--- via ffmpeg-cvslog wrote:
[...]
>     avformat/rawdec: set framerate in codec parameters
>     
>     Commit ba4b73c9779c32580f8a3ba08602a5d94e0bcd7c caused a regression in
>     the usage of avg_frame_rate to detect the frame rate of raw h264/hevc
>     bitstreams: after the commit, avg_frame_rate is always the value of the
>     -framerate option (which is set to 25 by default) instead of the actual
>     frame rate derived from the bitstream SPS/VPS NALUs.
>     
>     This commit fixes the regression by setting the framerate codec
>     parameter to the value of the framerate option instead. After this
>     change, bitstreams without timing information will derive avg_frame_rate
>     from the -framerate option, while bitstreams with timing information
>     will derive avg_frame_rate from the bitstream itself.
>
>     The h264-bsf-dts2pts test now returns the correct frame durations for a
>     bitstream with a mix of single-field and double-field frames.
>     
>     Signed-off-by: Gavin Li <git@thegavinli.com>
>     Signed-off-by: James Almer <jamrial@gmail.com>
> 
> diff --git a/libavformat/rawdec.c b/libavformat/rawdec.c
> index d0c829dc42..5cf2764a0d 100644
> --- a/libavformat/rawdec.c
> +++ b/libavformat/rawdec.c
> @@ -83,9 +83,9 @@ int ff_raw_video_read_header(AVFormatContext *s)
>  
>      st->codecpar->codec_type = AVMEDIA_TYPE_VIDEO;
>      st->codecpar->codec_id = ffifmt(s->iformat)->raw_codec_id;
> +    st->codecpar->framerate = s1->framerate;
>      sti->need_parsing = AVSTREAM_PARSE_FULL_RAW;
>  
> -    st->avg_frame_rate = s1->framerate;
>      avpriv_set_pts_info(st, 64, 1, 1200000);
>  
>  fail:
> diff --git a/tests/ref/fate/cavs-demux b/tests/ref/fate/cavs-demux
> index c4847293ab..eb16eb1f9d 100644
> --- a/tests/ref/fate/cavs-demux
> +++ b/tests/ref/fate/cavs-demux
> @@ -58,5 +58,5 @@ packet|codec_type=video|stream_index=0|pts=2240000|pts_time=1.866667|dts=2240000
>  packet|codec_type=video|stream_index=0|pts=2280000|pts_time=1.900000|dts=2280000|dts_time=1.900000|duration=40000|duration_time=0.033333|size=67|pos=172185|flags=___|data_hash=CRC32:42484449
>  packet|codec_type=video|stream_index=0|pts=2320000|pts_time=1.933333|dts=2320000|dts_time=1.933333|duration=40000|duration_time=0.033333|size=83|pos=172252|flags=___|data_hash=CRC32:a941bdf0
>  packet|codec_type=video|stream_index=0|pts=2360000|pts_time=1.966667|dts=2360000|dts_time=1.966667|duration=40000|duration_time=0.033333|size=5417|pos=172335|flags=___|data_hash=CRC32:9d0d503b
> -stream|index=0|codec_name=cavs|profile=unknown|codec_type=video|codec_tag_string=[0][0][0][0]|codec_tag=0x0000|width=1280|height=720|coded_width=1280|coded_height=720|has_b_frames=0|sample_aspect_ratio=N/A|display_aspect_ratio=N/A|pix_fmt=yuv420p|level=-99|color_range=unknown|color_space=unknown|color_transfer=unknown|color_primaries=unknown|chroma_location=unspecified|field_order=unknown|refs=1|id=N/A|r_frame_rate=30/1|avg_frame_rate=25/1|time_base=1/1200000|start_pts=N/A|start_time=N/A|duration_ts=N/A|duration=N/A|bit_rate=N/A|max_bit_rate=N/A|bits_per_raw_sample=N/A|nb_frames=N/A|nb_read_frames=N/A|nb_read_packets=60|extradata_size=18|extradata_hash=CRC32:1255d52e|disposition:default=0|disposition:dub=0|disposition:original=0|disposition:comment=0|disposition:lyrics=0|disposition:karaoke=0|disposition:forced=0|disposition:hearing_impaired=0|disposition:visual_impaired=0|disposition:clean_effects=0|disposition:attached_pic=0|disposition:timed_thumbnails=0|disposition:non_diegetic=
>  0|disposition:captions=0|disposition:descriptions=0|disposition:metadata=0|disposition:dependent=0|disposition:still_image=0|disposition:multilayer=0
> +stream|index=0|codec_name=cavs|profile=unknown|codec_type=video|codec_tag_string=[0][0][0][0]|codec_tag=0x0000|width=1280|height=720|coded_width=1280|coded_height=720|has_b_frames=0|sample_aspect_ratio=N/A|display_aspect_ratio=N/A|pix_fmt=yuv420p|level=-99|color_range=unknown|color_space=unknown|color_transfer=unknown|color_primaries=unknown|chroma_location=unspecified|field_order=unknown|refs=1|id=N/A|r_frame_rate=30/1|avg_frame_rate=30/1|time_base=1/1200000|start_pts=N/A|start_time=N/A|duration_ts=N/A|duration=N/A|bit_rate=N/A|max_bit_rate=N/A|bits_per_raw_sample=N/A|nb_frames=N/A|nb_read_frames=N/A|nb_read_packets=60|extradata_size=18|extradata_hash=CRC32:1255d52e|disposition:default=0|disposition:dub=0|disposition:original=0|disposition:comment=0|disposition:lyrics=0|disposition:karaoke=0|disposition:forced=0|disposition:hearing_impaired=0|disposition:visual_impaired=0|disposition:clean_effects=0|disposition:attached_pic=0|disposition:timed_thumbnails=0|disposition:non_diegetic=
>  0|disposition:captions=0|disposition:descriptions=0|disposition:metadata=0|disposition:dependent=0|disposition:still_image=0|disposition:multilayer=0
>  format|filename=bunny.mp4|nb_streams=1|nb_programs=0|nb_stream_groups=0|format_name=cavsvideo|start_time=N/A|duration=N/A|size=177752|bit_rate=N/A|probe_score=51

> diff --git a/tests/ref/fate/enhanced-flv-hevc-hdr10 b/tests/ref/fate/enhanced-flv-hevc-hdr10
> index 525f056d66..bebcf84fab 100644
> --- a/tests/ref/fate/enhanced-flv-hevc-hdr10
> +++ b/tests/ref/fate/enhanced-flv-hevc-hdr10
> @@ -4,7 +4,7 @@
>  #codec_id 0: hevc
>  #dimensions 0: 1280x720
>  #sar 0: 0/1
> -0,          0,          0,       40,    77718, 0xb59c83a5
> +0,          0,          0,        0,    77718, 0xb59c83a5
>  [FRAME]

The frame duration disappears here


[...]

> diff --git a/tests/ref/fate/h264-conformance-cvpcmnl2_sva_c b/tests/ref/fate/h264-conformance-cvpcmnl2_sva_c
> index 0303bc24e6..0a0795df41 100644
> --- a/tests/ref/fate/h264-conformance-cvpcmnl2_sva_c
> +++ b/tests/ref/fate/h264-conformance-cvpcmnl2_sva_c
> @@ -1,7 +1,7 @@
> -#tb 0: 1/25
> +#tb 0: 1/50
>  #media_type 0: video
>  #codec_id 0: rawvideo
>  #dimensions 0: 1280x720
>  #sar 0: 0/1
> -0,          0,          0,        1,  1382400, 0xccbe6bf8
> -0,          1,          1,        1,  1382400, 0x49c0cfd7
> +0,          0,          0,        2,  1382400, 0xccbe6bf8
> +0,          2,          2,        2,  1382400, 0x49c0cfd7

why is this (25fps) data becoming 1/50 ?


[...]
-- 
Michael     GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB

The smallest minority on earth is the individual. Those who deny 
individual rights cannot claim to be defenders of minorities. - Ayn Rand

[-- Attachment #1.2: signature.asc --]
[-- Type: application/pgp-signature, Size: 195 bytes --]

[-- Attachment #2: Type: text/plain, Size: 163 bytes --]

_______________________________________________
ffmpeg-devel mailing list -- ffmpeg-devel@ffmpeg.org
To unsubscribe send an email to ffmpeg-devel-leave@ffmpeg.org

[FFmpeg-devel] Re: [FFmpeg-cvslog] [ffmpeg] branch master updated. 92f2f9ea5c avcodec/d3d12va_encode: D3D12 AV1 encoding support

[Nicolas George via ffmpeg-devel]

Michael Niedermayer via ffmpeg-devel (HE12025-12-02):
> > +    assert((total_size + tile_group_extra_size) * 8 == bit_len);
> this possibly should be av_assert*

Without a doubt.

Michael Niedermayer via ffmpeg-devel (HE12025-12-02):
> this patch breaks decoding a video with a single image:
> 
> this works and creates our single image video
> ./ffmpeg -i lena.pnm /tmp/file.m2v
> 
> this fails after the patch:
> ./ffmpeg -i /tmp/file.m2v /tmp/file.jpg -y

This is bad.

Can we acknowledge that we observe a worsening in the scrutiny that
patches receive before being pushed?

Also, right now: make a pull request with a trivial patch, a
controversial one and another trivial patch, wait a few days, push:
nobody will notice because the PR will be named after the first,
trivial, patch, and the push notification will be named after the last,
trivial patch.

Regards,

-- 
  NicolasGeorge
_______________________________________________
ffmpeg-devel mailing list -- ffmpeg-devel@ffmpeg.org
To unsubscribe send an email to ffmpeg-devel-leave@ffmpeg.org

[FFmpeg-devel] Re: [FFmpeg-cvslog] [ffmpeg] branch master updated. 92f2f9ea5c avcodec/d3d12va_encode: D3D12 AV1 encoding support

[Michael Niedermayer via ffmpeg-devel]

[-- Attachment #1.1: Type: text/plain, Size: 1783 bytes --]

Hi Nicolas

On Tue, Dec 02, 2025 at 01:07:08PM +0100, Nicolas George via ffmpeg-devel wrote:
> Michael Niedermayer via ffmpeg-devel (HE12025-12-02):
> > > +    assert((total_size + tile_group_extra_size) * 8 == bit_len);
> > this possibly should be av_assert*
> 
> Without a doubt.
> 
> Michael Niedermayer via ffmpeg-devel (HE12025-12-02):
> > this patch breaks decoding a video with a single image:
> > 
> > this works and creates our single image video
> > ./ffmpeg -i lena.pnm /tmp/file.m2v
> > 
> > this fails after the patch:
> > ./ffmpeg -i /tmp/file.m2v /tmp/file.jpg -y
> 
> This is bad.
> 

> Can we acknowledge that we observe a worsening in the scrutiny that
> patches receive before being pushed?

This would require some statistics work, to be sure

In this specific case i simply didnt test it quick enough, i have a script
that warns me of new assert() so i should notice eventually.


> 
> Also, right now: make a pull request with a trivial patch, a
> controversial one and another trivial patch, wait a few days, push:
> nobody will notice because the PR will be named after the first,
> trivial, patch, and the push notification will be named after the last,
> trivial patch.

These are several seperate issues.

First, forgejo simply doesnt send the right details to ffmpeg-devel.

Second, our current "cvslog" reporting code is pure shit, i suggested previously to
switch this. nothing happened, ok, ill switch it, will try it with ffmpeg-web first

thx

[...]
-- 
Michael     GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB

Into a blind darkness they enter who follow after the Ignorance,
they as if into a greater darkness enter who devote themselves
to the Knowledge alone. -- Isha Upanishad

[-- Attachment #1.2: signature.asc --]
[-- Type: application/pgp-signature, Size: 195 bytes --]

[-- Attachment #2: Type: text/plain, Size: 163 bytes --]

_______________________________________________
ffmpeg-devel mailing list -- ffmpeg-devel@ffmpeg.org
To unsubscribe send an email to ffmpeg-devel-leave@ffmpeg.org

[FFmpeg-devel] Re: [FFmpeg-cvslog] [ffmpeg] branch master updated. 3d96d83a0a avformat/rawdec: set framerate in codec parameters

[James Almer via ffmpeg-devel]

[-- Attachment #1.1.1: Type: text/plain, Size: 2080 bytes --]

On 12/2/2025 12:36 AM, Michael Niedermayer via ffmpeg-devel wrote:
> Hi Gavin
> 
> this patch breaks decoding a video with a single image:
> 
> this works and creates our single image video
> ./ffmpeg -i lena.pnm /tmp/file.m2v
> 
> this fails after the patch:
> ./ffmpeg -i /tmp/file.m2v /tmp/file.jpg -y
> 
> [image2 @ 0x60dca024f140] Cannot write more than one file with the same name. Are you missing the -update option or a sequence pattern?
> [vost#0:0/mjpeg @ 0x60dca024f3c0] Error submitting a packet to the muxer: Invalid argument
>      Last message repeated 1 times
> [out#0/image2 @ 0x60dca024f040] Error muxing a packet
> [out#0/image2 @ 0x60dca024f040] Task finished with error code: -22 (Invalid argument)
> [out#0/image2 @ 0x60dca024f040] Terminating thread with return code -22 (Invalid argument)
> [out#0/image2 @ 0x60dca024f040] video:375KiB audio:0KiB subtitle:0KiB other streams:0KiB global headers:0KiB muxing overhead: unknown
> frame=    2 fps=0.0 q=5.0 Lsize=N/A time=00:00:00.04 bitrate=N/A dup=1 drop=0 speed=0.455x elapsed=0:00:00.08
> Conversion failed!
> 
> This also breaks:
> ./ffmpeg -i https://trac.ffmpeg.org/raw-attachment/ticket/3923/frame16.h264 test.jpg
> 
> rawdec has a user parameter called framerate which is set to 25 by default.
> If your input is not 25, obviously then it should not be set to 25.
> 
> Your change does not achive a fallback of 25 if it cannot be determined,
> as clearly shown in the example above where our single frame is turned into
> 2 frames, clearly showing there is a new inconsistancy.
> 
> I suggest to revert this patch (to avoid creating more bugs by a rushed attempt to
> fix this) and attempt to fix your file by setting the user framerate parameter correctly
> 
> Independant of this, making it autodetect would be cool, yes, i agree
> but i think we need more time, testing and no regressions

As the committer, I'm ok reverting it if it introduces regressions, but 
we should work towards having this fixed considering many of the fate 
changes were correct.


[-- Attachment #1.2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 495 bytes --]

[-- Attachment #2: Type: text/plain, Size: 163 bytes --]

_______________________________________________
ffmpeg-devel mailing list -- ffmpeg-devel@ffmpeg.org
To unsubscribe send an email to ffmpeg-devel-leave@ffmpeg.org