Re: [FFmpeg-devel] [PATCH v5 1/4] lavc/libdavs2: fix sequence incomplete output error
Any comments on these patches?
At 2018-11-02 21:30:08, "hwren" wrote:
>Signed-off-by: hwren
>---
> libavcodec/libdavs2.c | 12 +++-
> 1 file changed, 11 insertions(+), 1 deletion(-)
>
>diff --git a/libavcodec/libdavs2.c b/libavcodec/libdavs2.c
>index cadf995..e36bfed 100644
>--- a/libavcodec/libdavs2.c
>+++ b/libavcodec/libdavs2.c
>@@ -129,7 +129,17 @@ static int davs2_decode_frame(AVCodecContext *avctx, void
>*data,
> int ret = DAVS2_DEFAULT;
>
> if (!buf_size) {
>-return 0;
>+ret = davs2_decoder_flush(cad->decoder, >headerset,
>>out_frame);
>+if (ret == DAVS2_END) {
>+return 0;
>+} else if (ret == DAVS2_GOT_FRAME) {
>+*got_frame = davs2_dump_frames(avctx, >out_frame,
>>headerset, ret, frame);
>+davs2_decoder_frame_unref(cad->decoder, >out_frame);
>+return ret;
>+} else {
>+av_log(avctx, AV_LOG_ERROR, "Decoder error: dump frames
>failed\n");
>+return AVERROR_EXTERNAL;
>+}
> }
>
> cad->packet.data = buf_ptr;
>--
>2.7.4
___
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
[FFmpeg-devel] [PATCH] avcodec/cbs_av1: don't abort when splitting Temporal Units containing OBUs with no obu_size field
The ISOBMFF and Matroska specs allow the last OBU in a Sample/Block to have
obu_has_size_field equal to 0.
Signed-off-by: James Almer
---
See https://0x0.st/sUsU.mp4
It apparently can't be decoded with either aom or dav1d atm, but the latter
should be able to soon.
You can try remuxing it using the av1_metadata bsf after this patch while at it,
since the code to write obu_size fields unconditionally is still in place, and
that should make it decodable with either library.
For some reason cbs_av1_read_unit() is already able to handle OBUs with no
obu_size field. It was only cbs_av1_split_fragment() bailing out with them.
libavcodec/cbs_av1.c | 16 ++--
1 file changed, 6 insertions(+), 10 deletions(-)
diff --git a/libavcodec/cbs_av1.c b/libavcodec/cbs_av1.c
index ed5211be19..e02bc7027a 100644
--- a/libavcodec/cbs_av1.c
+++ b/libavcodec/cbs_av1.c
@@ -788,13 +788,6 @@ static int cbs_av1_split_fragment(CodedBitstreamContext
*ctx,
if (err < 0)
goto fail;
-if (!header.obu_has_size_field) {
-av_log(ctx->log_ctx, AV_LOG_ERROR, "Invalid OBU for raw "
- "stream: size field must be present.\n");
-err = AVERROR_INVALIDDATA;
-goto fail;
-}
-
if (get_bits_left() < 8) {
av_log(ctx->log_ctx, AV_LOG_ERROR, "Invalid OBU: fragment "
"too short (%zu bytes).\n", size);
@@ -802,9 +795,12 @@ static int cbs_av1_split_fragment(CodedBitstreamContext
*ctx,
goto fail;
}
-err = cbs_av1_read_leb128(ctx, , "obu_size", _size);
-if (err < 0)
-goto fail;
+if (header.obu_has_size_field) {
+err = cbs_av1_read_leb128(ctx, , "obu_size", _size);
+if (err < 0)
+goto fail;
+} else
+obu_size = size - 1 - header.obu_extension_flag;
pos = get_bits_count();
av_assert0(pos % 8 == 0 && pos / 8 <= size);
--
2.19.1
___
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
Re: [FFmpeg-devel] [PATCH 0/2] Update vf_bwdif to use yadif_common v2
On Tue, 13 Nov 2018 02:11:15 +0100 Thomas Mundt wrote: > Am So., 11. Nov. 2018 um 20:47 Uhr schrieb Philip Langdale < > phil...@overt.org>: > > > vf_bwdif's frame management logic is almost identical to that of > > yadif. The only difference is that it tracks the first and last > > fields in a sequence, and that requires slight changes to the > > common code. > > > > Assuming it's reasonable to do that tracking even though yadif > > doesn't need it, we can then remove all the duplicated logic. > > > > v2: Rename enum values as recommened by Thomas Mundt. > > > > Philip Langdale (2): > > avfilter/yadif_common: Add field type tracking to help bwdif > > avfilter/vf_bwdif: Use common yadif frame management logic > > > > libavfilter/bwdif.h | 34 + > > libavfilter/vf_bwdif.c | 235 > > +--- libavfilter/x86/vf_bwdif_init.c > > | 3 +- libavfilter/yadif.h | 14 ++ > > libavfilter/yadif_common.c | 12 +- > > 5 files changed, 64 insertions(+), 234 deletions(-) > > > > Patch lgtm. Thanks. Pushed. --phil ___ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
Re: [FFmpeg-devel] [PATCH] web/index: add news entry for release 4.1
On Wed, Nov 14, 2018 at 01:40:09PM -0300, James Almer wrote: > Signed-off-by: James Almer > --- > src/index | 50 ++ > 1 file changed, 50 insertions(+) LGTM thx [...] -- Michael GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB He who knows, does not speak. He who speaks, does not know. -- Lao Tsu signature.asc Description: PGP signature ___ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
Re: [FFmpeg-devel] [PATCH 2/2] avcodec/cbs_av1: fix parsing signed integer values
On 11/14/2018 8:04 PM, Mark Thompson wrote:
> On 14/11/18 22:39, James Almer wrote:
>> On 11/14/2018 7:24 PM, Mark Thompson wrote:
>>> On 11/11/18 02:24, James Almer wrote:
Signed-off-by: James Almer
---
See https://0x0.st/sljR.webm
libavcodec/cbs_av1.c | 30 +-
1 file changed, 9 insertions(+), 21 deletions(-)
diff --git a/libavcodec/cbs_av1.c b/libavcodec/cbs_av1.c
index ff32a6fca5..215f9384e8 100644
--- a/libavcodec/cbs_av1.c
+++ b/libavcodec/cbs_av1.c
@@ -189,30 +189,26 @@ static int cbs_av1_read_su(CodedBitstreamContext
*ctx, GetBitContext *gbc,
int width, const char *name,
const int *subscripts, int32_t *write_to)
{
-uint32_t magnitude;
-int position, sign;
+int position;
int32_t value;
if (ctx->trace_enable)
position = get_bits_count(gbc);
-if (get_bits_left(gbc) < width + 1) {
+if (get_bits_left(gbc) < width) {
av_log(ctx->log_ctx, AV_LOG_ERROR, "Invalid signed value at "
"%s: bitstream ended.\n", name);
return AVERROR_INVALIDDATA;
}
-magnitude = get_bits(gbc, width);
-sign = get_bits1(gbc);
-value = sign ? -(int32_t)magnitude : magnitude;
+value = get_sbits(gbc, width);
if (ctx->trace_enable) {
char bits[33];
int i;
for (i = 0; i < width; i++)
-bits[i] = magnitude >> (width - i - 1) & 1 ? '1' : '0';
-bits[i] = sign ? '1' : '0';
-bits[i + 1] = 0;
+bits[i] = value >> (width - i - 1) & 1 ? '1' : '0';
>>>
>>> Not sure if we care, but right-shifting a negative value is
>>> implementation-defined.
>>
>> Is casting value to unsigned enough, or should i use something like
>> zero_extend()?
>
> I think using left-shift instead is safe:
>
> value & 1 << (...) ? '1' : '0'
Changed and both patches applied and backported. Thanks!
___
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
Re: [FFmpeg-devel] [PATCH] [HLS] Add LANGUAGE attribute to #EXT-X-MEDIA tag for audio-only variant streams.
Hello everyone,
Here is the new version of the patch in which the comments on the curly
braces have been resolved as well.
I hope it's okay now.
Best regards,
Philippe Symons
Op do 15 nov. 2018 om 00:13 schreef Philippe Symons <
philippe.sym...@gmail.com>:
> Hold on, I just noticed the feedback on the curly braces. I'll create a
> new version of the patch to resolve this as well.
>
> Op wo 14 nov. 2018 om 23:49 schreef Philippe Symons <
> philippe.sym...@gmail.com>:
>
>> Hello,
>>
>> I've created a new version of the patch in which the second empty line
>> was removed. I couldn't find any other feedback inline in your response, so
>> that's all I changed.
>>
>> Best regards,
>>
>> Philippe Symons
>>
>>
>> Op di 13 nov. 2018 om 03:29 schreef Liu Steven :
>>
>>>
>>>
>>> > 在 2018年11月13日，上午6:36，Philippe Symons 写道：
>>> >
>>> > Hello everyone,
>>> >
>>> > This is my first patch to FFMpeg or any open-source project for that
>>> > matter. In this patch, I've added the LANGUAGE attribute to the
>>> > #EXT-X-MEDIA tag for for audio only variant streams for HLS streams.
>>> >
>>> > This attribute will be added to the output in
>>> > avformats' ff_hls_write_audio_rendition() function, if the language
>>> code
>>> > can be found in the metadata of the variant stream.
>>> >
>>> > The reason why I implemented this, was because I was working with a
>>> player
>>> > which needs this attribute to determine the language of the audio
>>> stream.
>>> > After asking the question on StackOverflow (
>>> >
>>> https://stackoverflow.com/questions/53205283/ffmpeg-hls-multiple-audio-languages-with-var-stream-map-ext-x-mediatype-doesn/53206492#53206492
>>> ),
>>> > I noticed that it wasn't implemented.
>>> >
>>> > I've tested these changes with the following command:
>>> >
>>> > ffmpeg -re -i /home/philippe/Videos/example.ts -map 0:v -c:v copy -map
>>> 0:a
>>> > -c:a copy -f hls -hls_time 6 -hls_list_size 10 -hls_flags
>>> > delete_segments+program_date_time -hls_segment_filename
>>> > "/var/www/html/live/stream_%v_%d.ts" -var_stream_map "v:0,agroup:lang
>>> > a:0,agroup:lang a:1,agroup:lang a:2,agroup:lang a:
>>> > 3,agroup:lang" -master_pl_name master.m3u8
>>> /var/www/html/live/stream-%v.m3u8
>>> >
>>> > Looking forward to your feedback. I hope it's okay.
>>> >
>>>
>>> Signed-off-by: Philippe Symons
>>> ---
>>> libavformat/dashenc.c | 2 +-
>>> libavformat/hlsenc.c | 24 ++--
>>> libavformat/hlsplaylist.c | 6 +-
>>> libavformat/hlsplaylist.h | 2 +-
>>> 4 files changed, 29 insertions(+), 5 deletions(-)
>>>
>>> diff --git a/libavformat/dashenc.c b/libavformat/dashenc.c
>>> index f8b3d106d5..ca5ef01e3f 100644
>>> --- a/libavformat/dashenc.c
>>> +++ b/libavformat/dashenc.c
>>> @@ -897,7 +897,7 @@ static int write_manifest(AVFormatContext *s, int
>>> final)
>>> if (st->codecpar->codec_type != AVMEDIA_TYPE_AUDIO)
>>> continue;
>>> get_hls_playlist_name(playlist_file, sizeof(playlist_file),
>>> NULL, i);
>>> -ff_hls_write_audio_rendition(c->m3u8_out, (char
>>> *)audio_group,
>>> +ff_hls_write_audio_rendition(c->m3u8_out, (char
>>> *)audio_group, NULL,
>>> playlist_file, i, is_default);
>>> max_audio_bitrate = FFMAX(st->codecpar->bit_rate +
>>>os->muxer_overhead,
>>> max_audio_bitrate);
>>> diff --git a/libavformat/hlsenc.c b/libavformat/hlsenc.c
>>> index e0048aa9d8..b31daa108e 100644
>>> --- a/libavformat/hlsenc.c
>>> +++ b/libavformat/hlsenc.c
>>> @@ -1205,7 +1205,6 @@ static int create_master_playlist(AVFormatContext
>>> *s,
>>> }
>>>
>>> set_http_options(s, , hls);
>>> -
>>> ret = hlsenc_io_open(s, >m3u8_out, hls->master_m3u8_url,
>>> );
>>> av_dict_free();
>>> if (ret < 0) {
>>> @@ -1228,12 +1227,33 @@ static int
>>> create_master_playlist(AVFormatContext *s,
>>>
>>> /* For audio only variant streams add #EXT-X-MEDIA tag with
>>> attributes*/
>>> for (i = 0; i < hls->nb_varstreams; i++) {
>>> +AVFormatContext* var_context;
>>> +char* language = NULL;
>>> +
>>> vs = &(hls->var_streams[i]);
>>> +var_context = vs->avf;
>>>
>>> if (vs->has_video || vs->has_subtitle || !vs->agroup)
>>> continue;
>>>
>>> m3u8_name_size = strlen(vs->m3u8_name) + 1;
>>> +
>>> +/*
>>> + * Try to obtain the language code of the audio stream.
>>> + * -if available- it will be used to write the LANGUAGE
>>> + * attribute in the #EXT-X-MEDIA tag
>>> + */
>>> +if(var_context && var_context->nb_streams > 0)
>>> +{
>>> if (var_context && var_context->nb_streams > 0) {
>>> +AVDictionary* meta_dict = vs->streams[0]->metadata;
>>> +AVDictionaryEntry *langEntry = av_dict_get(meta_dict,
>>> "language", NULL, 0);
>>> +if(langEntry)
>>> +{
>>> if (langEntry) {
>>> +
Re: [FFmpeg-devel] [PATCH] configure: Add missing xlib dependency for VAAPI X11 code
On 14/11/18 23:02, James Almer wrote:
> On 11/14/2018 7:59 PM, Mark Thompson wrote:
>> Fixes #7538.
>> ---
>> $ ../configure --disable-vdpau --enable-vaapi
>> ...
>> $ rm libavutil/libavutil.pc ; make libavutil/libavutil.pc ; cat
>> libavutil/libavutil.pc
>> ...
>>
>> Before:
>> Libs: -L${libdir} -lavutil -pthread -L/usr/local/lib -lva-drm -lva
>> -L/usr/local/lib -lva-x11 -lva -lm -L/usr/local/lib -lva
>>
>> After:
>> Libs: -L${libdir} -lavutil -pthread -L/usr/local/lib -lva-drm -lva
>> -L/usr/local/lib -lva-x11 -lva -lm -L/usr/local/lib -lva -lXv -lX11 -lXext
>>
>>
>> configure | 1 +
>> 1 file changed, 1 insertion(+)
>>
>> diff --git a/configure b/configure
>> index b02b4ccb2e..9bc4cf330d 100755
>> --- a/configure
>> +++ b/configure
>> @@ -2815,6 +2815,7 @@ d3d11va_deps="dxva_h ID3D11VideoDecoder
>> ID3D11VideoContext"
>> dxva2_deps="dxva2api_h DXVA2_ConfigPictureDecode ole32 user32"
>> ffnvcodec_deps_any="libdl LoadLibrary"
>> nvdec_deps="ffnvcodec"
>> +vaapi_x11_deps="xlib"
>> videotoolbox_hwaccel_deps="videotoolbox pthreads"
>> videotoolbox_hwaccel_extralibs="-framework QuartzCore"
>> xvmc_deps="X11_extensions_XvMClib_h"
>>
>
> LGTM.
>
> Opening a ticket in libva's repo so they fix their pkg-config files may
> be worth it as well.
I'm not sure there is anything which can usefully be fixed in libva:
* The shared libraries themselves have the right internal linking (DT_NEEDED),
so they don't need anything extra at link time.
* Static linking libva isn't really supported at all - libva depends on dynamic
linking, including being able to find its own symbols backwards from a
dynamically-loaded driver binary. (It's probably possible by some method
involving --export-dynamic, though I haven't tried.)
Anyway, applied. I'll backport to 4.1 as well, since that was where the
original report came from.
Thanks,
- Mark
___
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
Re: [FFmpeg-devel] [PATCH] [HLS] Add LANGUAGE attribute to #EXT-X-MEDIA tag for audio-only variant streams.
Hold on, I just noticed the feedback on the curly braces. I'll create a new
version of the patch to resolve this as well.
Op wo 14 nov. 2018 om 23:49 schreef Philippe Symons <
philippe.sym...@gmail.com>:
> Hello,
>
> I've created a new version of the patch in which the second empty line was
> removed. I couldn't find any other feedback inline in your response, so
> that's all I changed.
>
> Best regards,
>
> Philippe Symons
>
>
> Op di 13 nov. 2018 om 03:29 schreef Liu Steven :
>
>>
>>
>> > 在 2018年11月13日，上午6:36，Philippe Symons 写道：
>> >
>> > Hello everyone,
>> >
>> > This is my first patch to FFMpeg or any open-source project for that
>> > matter. In this patch, I've added the LANGUAGE attribute to the
>> > #EXT-X-MEDIA tag for for audio only variant streams for HLS streams.
>> >
>> > This attribute will be added to the output in
>> > avformats' ff_hls_write_audio_rendition() function, if the language code
>> > can be found in the metadata of the variant stream.
>> >
>> > The reason why I implemented this, was because I was working with a
>> player
>> > which needs this attribute to determine the language of the audio
>> stream.
>> > After asking the question on StackOverflow (
>> >
>> https://stackoverflow.com/questions/53205283/ffmpeg-hls-multiple-audio-languages-with-var-stream-map-ext-x-mediatype-doesn/53206492#53206492
>> ),
>> > I noticed that it wasn't implemented.
>> >
>> > I've tested these changes with the following command:
>> >
>> > ffmpeg -re -i /home/philippe/Videos/example.ts -map 0:v -c:v copy -map
>> 0:a
>> > -c:a copy -f hls -hls_time 6 -hls_list_size 10 -hls_flags
>> > delete_segments+program_date_time -hls_segment_filename
>> > "/var/www/html/live/stream_%v_%d.ts" -var_stream_map "v:0,agroup:lang
>> > a:0,agroup:lang a:1,agroup:lang a:2,agroup:lang a:
>> > 3,agroup:lang" -master_pl_name master.m3u8
>> /var/www/html/live/stream-%v.m3u8
>> >
>> > Looking forward to your feedback. I hope it's okay.
>> >
>>
>> Signed-off-by: Philippe Symons
>> ---
>> libavformat/dashenc.c | 2 +-
>> libavformat/hlsenc.c | 24 ++--
>> libavformat/hlsplaylist.c | 6 +-
>> libavformat/hlsplaylist.h | 2 +-
>> 4 files changed, 29 insertions(+), 5 deletions(-)
>>
>> diff --git a/libavformat/dashenc.c b/libavformat/dashenc.c
>> index f8b3d106d5..ca5ef01e3f 100644
>> --- a/libavformat/dashenc.c
>> +++ b/libavformat/dashenc.c
>> @@ -897,7 +897,7 @@ static int write_manifest(AVFormatContext *s, int
>> final)
>> if (st->codecpar->codec_type != AVMEDIA_TYPE_AUDIO)
>> continue;
>> get_hls_playlist_name(playlist_file, sizeof(playlist_file),
>> NULL, i);
>> -ff_hls_write_audio_rendition(c->m3u8_out, (char
>> *)audio_group,
>> +ff_hls_write_audio_rendition(c->m3u8_out, (char
>> *)audio_group, NULL,
>> playlist_file, i, is_default);
>> max_audio_bitrate = FFMAX(st->codecpar->bit_rate +
>>os->muxer_overhead,
>> max_audio_bitrate);
>> diff --git a/libavformat/hlsenc.c b/libavformat/hlsenc.c
>> index e0048aa9d8..b31daa108e 100644
>> --- a/libavformat/hlsenc.c
>> +++ b/libavformat/hlsenc.c
>> @@ -1205,7 +1205,6 @@ static int create_master_playlist(AVFormatContext
>> *s,
>> }
>>
>> set_http_options(s, , hls);
>> -
>> ret = hlsenc_io_open(s, >m3u8_out, hls->master_m3u8_url,
>> );
>> av_dict_free();
>> if (ret < 0) {
>> @@ -1228,12 +1227,33 @@ static int create_master_playlist(AVFormatContext
>> *s,
>>
>> /* For audio only variant streams add #EXT-X-MEDIA tag with
>> attributes*/
>> for (i = 0; i < hls->nb_varstreams; i++) {
>> +AVFormatContext* var_context;
>> +char* language = NULL;
>> +
>> vs = &(hls->var_streams[i]);
>> +var_context = vs->avf;
>>
>> if (vs->has_video || vs->has_subtitle || !vs->agroup)
>> continue;
>>
>> m3u8_name_size = strlen(vs->m3u8_name) + 1;
>> +
>> +/*
>> + * Try to obtain the language code of the audio stream.
>> + * -if available- it will be used to write the LANGUAGE
>> + * attribute in the #EXT-X-MEDIA tag
>> + */
>> +if(var_context && var_context->nb_streams > 0)
>> +{
>> if (var_context && var_context->nb_streams > 0) {
>> +AVDictionary* meta_dict = vs->streams[0]->metadata;
>> +AVDictionaryEntry *langEntry = av_dict_get(meta_dict,
>> "language", NULL, 0);
>> +if(langEntry)
>> +{
>> if (langEntry) {
>> +language = langEntry->value;
>> +}
>> +}
>> +
>> +
>>
>> two empty line?
>> m3u8_rel_name = av_malloc(m3u8_name_size);
>> if (!m3u8_rel_name) {
>> ret = AVERROR(ENOMEM);
>> @@ -1247,7 +1267,7 @@ static int create_master_playlist(AVFormatContext
>> *s,
>> goto fail;
>> }
>>
>> -
Re: [FFmpeg-devel] [PATCH 2/2] avcodec/cbs_av1: fix parsing signed integer values
On 14/11/18 22:39, James Almer wrote:
> On 11/14/2018 7:24 PM, Mark Thompson wrote:
>> On 11/11/18 02:24, James Almer wrote:
>>> Signed-off-by: James Almer
>>> ---
>>> See https://0x0.st/sljR.webm
>>>
>>> libavcodec/cbs_av1.c | 30 +-
>>> 1 file changed, 9 insertions(+), 21 deletions(-)
>>>
>>> diff --git a/libavcodec/cbs_av1.c b/libavcodec/cbs_av1.c
>>> index ff32a6fca5..215f9384e8 100644
>>> --- a/libavcodec/cbs_av1.c
>>> +++ b/libavcodec/cbs_av1.c
>>> @@ -189,30 +189,26 @@ static int cbs_av1_read_su(CodedBitstreamContext
>>> *ctx, GetBitContext *gbc,
>>> int width, const char *name,
>>> const int *subscripts, int32_t *write_to)
>>> {
>>> -uint32_t magnitude;
>>> -int position, sign;
>>> +int position;
>>> int32_t value;
>>>
>>> if (ctx->trace_enable)
>>> position = get_bits_count(gbc);
>>>
>>> -if (get_bits_left(gbc) < width + 1) {
>>> +if (get_bits_left(gbc) < width) {
>>> av_log(ctx->log_ctx, AV_LOG_ERROR, "Invalid signed value at "
>>> "%s: bitstream ended.\n", name);
>>> return AVERROR_INVALIDDATA;
>>> }
>>>
>>> -magnitude = get_bits(gbc, width);
>>> -sign = get_bits1(gbc);
>>> -value = sign ? -(int32_t)magnitude : magnitude;
>>> +value = get_sbits(gbc, width);
>>>
>>> if (ctx->trace_enable) {
>>> char bits[33];
>>> int i;
>>> for (i = 0; i < width; i++)
>>> -bits[i] = magnitude >> (width - i - 1) & 1 ? '1' : '0';
>>> -bits[i] = sign ? '1' : '0';
>>> -bits[i + 1] = 0;
>>> +bits[i] = value >> (width - i - 1) & 1 ? '1' : '0';
>>
>> Not sure if we care, but right-shifting a negative value is
>> implementation-defined.
>
> Is casting value to unsigned enough, or should i use something like
> zero_extend()?
I think using left-shift instead is safe:
value & 1 << (...) ? '1' : '0'
___
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
Re: [FFmpeg-devel] [PATCH] configure: Add missing xlib dependency for VAAPI X11 code
On 11/14/2018 7:59 PM, Mark Thompson wrote:
> Fixes #7538.
> ---
> $ ../configure --disable-vdpau --enable-vaapi
> ...
> $ rm libavutil/libavutil.pc ; make libavutil/libavutil.pc ; cat
> libavutil/libavutil.pc
> ...
>
> Before:
> Libs: -L${libdir} -lavutil -pthread -L/usr/local/lib -lva-drm -lva
> -L/usr/local/lib -lva-x11 -lva -lm -L/usr/local/lib -lva
>
> After:
> Libs: -L${libdir} -lavutil -pthread -L/usr/local/lib -lva-drm -lva
> -L/usr/local/lib -lva-x11 -lva -lm -L/usr/local/lib -lva -lXv -lX11 -lXext
>
>
> configure | 1 +
> 1 file changed, 1 insertion(+)
>
> diff --git a/configure b/configure
> index b02b4ccb2e..9bc4cf330d 100755
> --- a/configure
> +++ b/configure
> @@ -2815,6 +2815,7 @@ d3d11va_deps="dxva_h ID3D11VideoDecoder
> ID3D11VideoContext"
> dxva2_deps="dxva2api_h DXVA2_ConfigPictureDecode ole32 user32"
> ffnvcodec_deps_any="libdl LoadLibrary"
> nvdec_deps="ffnvcodec"
> +vaapi_x11_deps="xlib"
> videotoolbox_hwaccel_deps="videotoolbox pthreads"
> videotoolbox_hwaccel_extralibs="-framework QuartzCore"
> xvmc_deps="X11_extensions_XvMClib_h"
>
LGTM.
Opening a ticket in libva's repo so they fix their pkg-config files may
be worth it as well.
___
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
[FFmpeg-devel] [PATCH] configure: Add missing xlib dependency for VAAPI X11 code
Fixes #7538.
---
$ ../configure --disable-vdpau --enable-vaapi
...
$ rm libavutil/libavutil.pc ; make libavutil/libavutil.pc ; cat
libavutil/libavutil.pc
...
Before:
Libs: -L${libdir} -lavutil -pthread -L/usr/local/lib -lva-drm -lva
-L/usr/local/lib -lva-x11 -lva -lm -L/usr/local/lib -lva
After:
Libs: -L${libdir} -lavutil -pthread -L/usr/local/lib -lva-drm -lva
-L/usr/local/lib -lva-x11 -lva -lm -L/usr/local/lib -lva -lXv -lX11 -lXext
configure | 1 +
1 file changed, 1 insertion(+)
diff --git a/configure b/configure
index b02b4ccb2e..9bc4cf330d 100755
--- a/configure
+++ b/configure
@@ -2815,6 +2815,7 @@ d3d11va_deps="dxva_h ID3D11VideoDecoder
ID3D11VideoContext"
dxva2_deps="dxva2api_h DXVA2_ConfigPictureDecode ole32 user32"
ffnvcodec_deps_any="libdl LoadLibrary"
nvdec_deps="ffnvcodec"
+vaapi_x11_deps="xlib"
videotoolbox_hwaccel_deps="videotoolbox pthreads"
videotoolbox_hwaccel_extralibs="-framework QuartzCore"
xvmc_deps="X11_extensions_XvMClib_h"
--
2.19.1
___
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
Re: [FFmpeg-devel] [PATCH] [HLS] Add LANGUAGE attribute to #EXT-X-MEDIA tag for audio-only variant streams.
Hello,
I've created a new version of the patch in which the second empty line was
removed. I couldn't find any other feedback inline in your response, so
that's all I changed.
Best regards,
Philippe Symons
Op di 13 nov. 2018 om 03:29 schreef Liu Steven :
>
>
> > 在 2018年11月13日，上午6:36，Philippe Symons 写道：
> >
> > Hello everyone,
> >
> > This is my first patch to FFMpeg or any open-source project for that
> > matter. In this patch, I've added the LANGUAGE attribute to the
> > #EXT-X-MEDIA tag for for audio only variant streams for HLS streams.
> >
> > This attribute will be added to the output in
> > avformats' ff_hls_write_audio_rendition() function, if the language code
> > can be found in the metadata of the variant stream.
> >
> > The reason why I implemented this, was because I was working with a
> player
> > which needs this attribute to determine the language of the audio stream.
> > After asking the question on StackOverflow (
> >
> https://stackoverflow.com/questions/53205283/ffmpeg-hls-multiple-audio-languages-with-var-stream-map-ext-x-mediatype-doesn/53206492#53206492
> ),
> > I noticed that it wasn't implemented.
> >
> > I've tested these changes with the following command:
> >
> > ffmpeg -re -i /home/philippe/Videos/example.ts -map 0:v -c:v copy -map
> 0:a
> > -c:a copy -f hls -hls_time 6 -hls_list_size 10 -hls_flags
> > delete_segments+program_date_time -hls_segment_filename
> > "/var/www/html/live/stream_%v_%d.ts" -var_stream_map "v:0,agroup:lang
> > a:0,agroup:lang a:1,agroup:lang a:2,agroup:lang a:
> > 3,agroup:lang" -master_pl_name master.m3u8
> /var/www/html/live/stream-%v.m3u8
> >
> > Looking forward to your feedback. I hope it's okay.
> >
>
> Signed-off-by: Philippe Symons
> ---
> libavformat/dashenc.c | 2 +-
> libavformat/hlsenc.c | 24 ++--
> libavformat/hlsplaylist.c | 6 +-
> libavformat/hlsplaylist.h | 2 +-
> 4 files changed, 29 insertions(+), 5 deletions(-)
>
> diff --git a/libavformat/dashenc.c b/libavformat/dashenc.c
> index f8b3d106d5..ca5ef01e3f 100644
> --- a/libavformat/dashenc.c
> +++ b/libavformat/dashenc.c
> @@ -897,7 +897,7 @@ static int write_manifest(AVFormatContext *s, int
> final)
> if (st->codecpar->codec_type != AVMEDIA_TYPE_AUDIO)
> continue;
> get_hls_playlist_name(playlist_file, sizeof(playlist_file),
> NULL, i);
> -ff_hls_write_audio_rendition(c->m3u8_out, (char *)audio_group,
> +ff_hls_write_audio_rendition(c->m3u8_out, (char
> *)audio_group, NULL,
> playlist_file, i, is_default);
> max_audio_bitrate = FFMAX(st->codecpar->bit_rate +
>os->muxer_overhead,
> max_audio_bitrate);
> diff --git a/libavformat/hlsenc.c b/libavformat/hlsenc.c
> index e0048aa9d8..b31daa108e 100644
> --- a/libavformat/hlsenc.c
> +++ b/libavformat/hlsenc.c
> @@ -1205,7 +1205,6 @@ static int create_master_playlist(AVFormatContext *s,
> }
>
> set_http_options(s, , hls);
> -
> ret = hlsenc_io_open(s, >m3u8_out, hls->master_m3u8_url,
> );
> av_dict_free();
> if (ret < 0) {
> @@ -1228,12 +1227,33 @@ static int create_master_playlist(AVFormatContext
> *s,
>
> /* For audio only variant streams add #EXT-X-MEDIA tag with
> attributes*/
> for (i = 0; i < hls->nb_varstreams; i++) {
> +AVFormatContext* var_context;
> +char* language = NULL;
> +
> vs = &(hls->var_streams[i]);
> +var_context = vs->avf;
>
> if (vs->has_video || vs->has_subtitle || !vs->agroup)
> continue;
>
> m3u8_name_size = strlen(vs->m3u8_name) + 1;
> +
> +/*
> + * Try to obtain the language code of the audio stream.
> + * -if available- it will be used to write the LANGUAGE
> + * attribute in the #EXT-X-MEDIA tag
> + */
> +if(var_context && var_context->nb_streams > 0)
> +{
> if (var_context && var_context->nb_streams > 0) {
> +AVDictionary* meta_dict = vs->streams[0]->metadata;
> +AVDictionaryEntry *langEntry = av_dict_get(meta_dict,
> "language", NULL, 0);
> +if(langEntry)
> +{
> if (langEntry) {
> +language = langEntry->value;
> +}
> +}
> +
> +
>
> two empty line?
> m3u8_rel_name = av_malloc(m3u8_name_size);
> if (!m3u8_rel_name) {
> ret = AVERROR(ENOMEM);
> @@ -1247,7 +1267,7 @@ static int create_master_playlist(AVFormatContext *s,
> goto fail;
> }
>
> -ff_hls_write_audio_rendition(hls->m3u8_out, vs->agroup,
> m3u8_rel_name, 0, 1);
> +ff_hls_write_audio_rendition(hls->m3u8_out, vs->agroup, language,
> m3u8_rel_name, 0, 1);
>
> av_freep(_rel_name);
> }
> diff --git a/libavformat/hlsplaylist.c b/libavformat/hlsplaylist.c
> index efcbff0009..42c7cf5727 100644
> --- a/libavformat/hlsplaylist.c
> +++
Re: [FFmpeg-devel] [PATCH 2/2] avcodec/cbs_av1: fix parsing signed integer values
On 11/14/2018 7:24 PM, Mark Thompson wrote:
> On 11/11/18 02:24, James Almer wrote:
>> Signed-off-by: James Almer
>> ---
>> See https://0x0.st/sljR.webm
>>
>> libavcodec/cbs_av1.c | 30 +-
>> 1 file changed, 9 insertions(+), 21 deletions(-)
>>
>> diff --git a/libavcodec/cbs_av1.c b/libavcodec/cbs_av1.c
>> index ff32a6fca5..215f9384e8 100644
>> --- a/libavcodec/cbs_av1.c
>> +++ b/libavcodec/cbs_av1.c
>> @@ -189,30 +189,26 @@ static int cbs_av1_read_su(CodedBitstreamContext *ctx,
>> GetBitContext *gbc,
>> int width, const char *name,
>> const int *subscripts, int32_t *write_to)
>> {
>> -uint32_t magnitude;
>> -int position, sign;
>> +int position;
>> int32_t value;
>>
>> if (ctx->trace_enable)
>> position = get_bits_count(gbc);
>>
>> -if (get_bits_left(gbc) < width + 1) {
>> +if (get_bits_left(gbc) < width) {
>> av_log(ctx->log_ctx, AV_LOG_ERROR, "Invalid signed value at "
>> "%s: bitstream ended.\n", name);
>> return AVERROR_INVALIDDATA;
>> }
>>
>> -magnitude = get_bits(gbc, width);
>> -sign = get_bits1(gbc);
>> -value = sign ? -(int32_t)magnitude : magnitude;
>> +value = get_sbits(gbc, width);
>>
>> if (ctx->trace_enable) {
>> char bits[33];
>> int i;
>> for (i = 0; i < width; i++)
>> -bits[i] = magnitude >> (width - i - 1) & 1 ? '1' : '0';
>> -bits[i] = sign ? '1' : '0';
>> -bits[i + 1] = 0;
>> +bits[i] = value >> (width - i - 1) & 1 ? '1' : '0';
>
> Not sure if we care, but right-shifting a negative value is
> implementation-defined.
Is casting value to unsigned enough, or should i use something like
zero_extend()?
>
>> +bits[i] = 0;
>>
>> ff_cbs_trace_syntax_element(ctx, position,
>> name, subscripts, bits, value);
>> @@ -226,29 +222,21 @@ static int cbs_av1_write_su(CodedBitstreamContext
>> *ctx, PutBitContext *pbc,
>> int width, const char *name,
>> const int *subscripts, int32_t value)
>> {
>> -uint32_t magnitude;
>> -int sign;
>> -
>> -if (put_bits_left(pbc) < width + 1)
>> +if (put_bits_left(pbc) < width)
>> return AVERROR(ENOSPC);
>>
>> -sign = value < 0;
>> -magnitude = sign ? -value : value;
>> -
>> if (ctx->trace_enable) {
>> char bits[33];
>> int i;
>> for (i = 0; i < width; i++)
>> -bits[i] = magnitude >> (width - i - 1) & 1 ? '1' : '0';
>> -bits[i] = sign ? '1' : '0';
>> -bits[i + 1] = 0;
>> +bits[i] = value >> (width - i - 1) & 1 ? '1' : '0';
>
> And here.
>
>> +bits[i] = 0;
>>
>> ff_cbs_trace_syntax_element(ctx, put_bits_count(pbc),
>> name, subscripts, bits, value);
>> }
>>
>> -put_bits(pbc, width, magnitude);
>> -put_bits(pbc, 1, sign);
>> +put_sbits(pbc, width, value);
>>
>> return 0;
>> }
>>
>
> Otherwise fine. No idea where the version with the sign bit at the other end
> came from; oh well.
From a lack of samples to test and confirm that the code works :p
>
> Thanks,
>
> - Mark
> ___
> ffmpeg-devel mailing list
> ffmpeg-devel@ffmpeg.org
> http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
>
___
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
Re: [FFmpeg-devel] [PATCH 2/2] avcodec/cbs_av1: fix parsing signed integer values
On 11/11/18 02:24, James Almer wrote:
> Signed-off-by: James Almer
> ---
> See https://0x0.st/sljR.webm
>
> libavcodec/cbs_av1.c | 30 +-
> 1 file changed, 9 insertions(+), 21 deletions(-)
>
> diff --git a/libavcodec/cbs_av1.c b/libavcodec/cbs_av1.c
> index ff32a6fca5..215f9384e8 100644
> --- a/libavcodec/cbs_av1.c
> +++ b/libavcodec/cbs_av1.c
> @@ -189,30 +189,26 @@ static int cbs_av1_read_su(CodedBitstreamContext *ctx,
> GetBitContext *gbc,
> int width, const char *name,
> const int *subscripts, int32_t *write_to)
> {
> -uint32_t magnitude;
> -int position, sign;
> +int position;
> int32_t value;
>
> if (ctx->trace_enable)
> position = get_bits_count(gbc);
>
> -if (get_bits_left(gbc) < width + 1) {
> +if (get_bits_left(gbc) < width) {
> av_log(ctx->log_ctx, AV_LOG_ERROR, "Invalid signed value at "
> "%s: bitstream ended.\n", name);
> return AVERROR_INVALIDDATA;
> }
>
> -magnitude = get_bits(gbc, width);
> -sign = get_bits1(gbc);
> -value = sign ? -(int32_t)magnitude : magnitude;
> +value = get_sbits(gbc, width);
>
> if (ctx->trace_enable) {
> char bits[33];
> int i;
> for (i = 0; i < width; i++)
> -bits[i] = magnitude >> (width - i - 1) & 1 ? '1' : '0';
> -bits[i] = sign ? '1' : '0';
> -bits[i + 1] = 0;
> +bits[i] = value >> (width - i - 1) & 1 ? '1' : '0';
Not sure if we care, but right-shifting a negative value is
implementation-defined.
> +bits[i] = 0;
>
> ff_cbs_trace_syntax_element(ctx, position,
> name, subscripts, bits, value);
> @@ -226,29 +222,21 @@ static int cbs_av1_write_su(CodedBitstreamContext *ctx,
> PutBitContext *pbc,
> int width, const char *name,
> const int *subscripts, int32_t value)
> {
> -uint32_t magnitude;
> -int sign;
> -
> -if (put_bits_left(pbc) < width + 1)
> +if (put_bits_left(pbc) < width)
> return AVERROR(ENOSPC);
>
> -sign = value < 0;
> -magnitude = sign ? -value : value;
> -
> if (ctx->trace_enable) {
> char bits[33];
> int i;
> for (i = 0; i < width; i++)
> -bits[i] = magnitude >> (width - i - 1) & 1 ? '1' : '0';
> -bits[i] = sign ? '1' : '0';
> -bits[i + 1] = 0;
> +bits[i] = value >> (width - i - 1) & 1 ? '1' : '0';
And here.
> +bits[i] = 0;
>
> ff_cbs_trace_syntax_element(ctx, put_bits_count(pbc),
> name, subscripts, bits, value);
> }
>
> -put_bits(pbc, width, magnitude);
> -put_bits(pbc, 1, sign);
> +put_sbits(pbc, width, value);
>
> return 0;
> }
>
Otherwise fine. No idea where the version with the sign bit at the other end
came from; oh well.
Thanks,
- Mark
___
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
Re: [FFmpeg-devel] [PATCH v2] lavf/isom: add Dolby Vision sample entry codes for HEVC and H.264
2018-11-14 20:59 GMT+01:00, Jan Ekström : > So what if I just make a patch without dvh1 that you have > already implemented, and then we call that a day? I thought I suggested that but I may misremember. Sorry, Carl Eugen ___ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
Re: [FFmpeg-devel] [PATCH 1/2] avcodec/cbs_av1: fix storage size for segmentation_params feature_value fields
On 11/11/18 02:24, James Almer wrote:
> The valid range is -255 to 255.
>
> Signed-off-by: James Almer
> ---
> libavcodec/cbs_av1.h | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/libavcodec/cbs_av1.h b/libavcodec/cbs_av1.h
> index f662265f75..84622ed189 100644
> --- a/libavcodec/cbs_av1.h
> +++ b/libavcodec/cbs_av1.h
> @@ -210,7 +210,7 @@ typedef struct AV1RawFrameHeader {
> uint8_t segmentation_temporal_update;
> uint8_t segmentation_update_data;
> uint8_t feature_enabled[AV1_MAX_SEGMENTS][AV1_SEG_LVL_MAX];
> -uint8_t feature_value[AV1_MAX_SEGMENTS][AV1_SEG_LVL_MAX];
> +int16_t feature_value[AV1_MAX_SEGMENTS][AV1_SEG_LVL_MAX];
>
> uint8_t delta_q_present;
> uint8_t delta_q_res;
>
Yep, LGTM.
Thanks,
- Mark
___
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
[FFmpeg-devel] [PATCH] hevcdec: Fix get_format behaviour when the user declines to pick a format
Maintain the context pixfmt as NONE in this case, indicating that if the
user requests more decoding then get_format() will need to be called again.
Also remove the pixfmt setting from inside set_sps() and make it explicit,
so that the behaviour is clearer.
---
libavcodec/hevcdec.c | 28 +---
1 file changed, 17 insertions(+), 11 deletions(-)
diff --git a/libavcodec/hevcdec.c b/libavcodec/hevcdec.c
index a3b5c8cb71..dca478db6c 100644
--- a/libavcodec/hevcdec.c
+++ b/libavcodec/hevcdec.c
@@ -421,8 +421,7 @@ static enum AVPixelFormat get_format(HEVCContext *s, const
HEVCSPS *sps)
return ff_thread_get_format(s->avctx, pix_fmts);
}
-static int set_sps(HEVCContext *s, const HEVCSPS *sps,
- enum AVPixelFormat pix_fmt)
+static int set_sps(HEVCContext *s, const HEVCSPS *sps)
{
int ret, i;
@@ -439,8 +438,6 @@ static int set_sps(HEVCContext *s, const HEVCSPS *sps,
export_stream_params(s->avctx, >ps, sps);
-s->avctx->pix_fmt = pix_fmt;
-
ff_hevc_pred_init(>hpc, sps->bit_depth);
ff_hevc_dsp_init (>hevcdsp, sps->bit_depth);
ff_videodsp_init (>vdsp,sps->bit_depth);
@@ -512,7 +509,6 @@ static int hls_slice_header(HEVCContext *s)
if (s->ps.sps != (HEVCSPS*)s->ps.sps_list[s->ps.pps->sps_id]->data) {
const HEVCSPS *sps = (HEVCSPS*)s->ps.sps_list[s->ps.pps->sps_id]->data;
const HEVCSPS *last_sps = s->ps.sps;
-enum AVPixelFormat pix_fmt;
if (last_sps && IS_IRAP(s) && s->nal_unit_type != HEVC_NAL_CRA_NUT) {
if (sps->width != last_sps->width || sps->height !=
last_sps->height ||
@@ -522,17 +518,25 @@ static int hls_slice_header(HEVCContext *s)
}
ff_hevc_clear_refs(s);
-ret = set_sps(s, sps, sps->pix_fmt);
+ret = set_sps(s, sps);
if (ret < 0)
return ret;
+s->seq_decode = (s->seq_decode + 1) & 0xff;
+s->max_ra = INT_MAX;
+
+s->avctx->pix_fmt = AV_PIX_FMT_NONE;
+}
+
+if (s->avctx->pix_fmt == AV_PIX_FMT_NONE) {
+const HEVCSPS *sps = (HEVCSPS*)s->ps.sps_list[s->ps.pps->sps_id]->data;
+enum AVPixelFormat pix_fmt;
+
pix_fmt = get_format(s, sps);
if (pix_fmt < 0)
-return pix_fmt;
-s->avctx->pix_fmt = pix_fmt;
+return AVERROR(EINVAL);
-s->seq_decode = (s->seq_decode + 1) & 0xff;
-s->max_ra = INT_MAX;
+s->avctx->pix_fmt = pix_fmt;
}
sh->dependent_slice_segment_flag = 0;
@@ -3411,9 +3415,11 @@ static int hevc_update_thread_context(AVCodecContext
*dst,
}
if (s->ps.sps != s0->ps.sps)
-if ((ret = set_sps(s, s0->ps.sps, src->pix_fmt)) < 0)
+if ((ret = set_sps(s, s0->ps.sps)) < 0)
return ret;
+dst->pix_fmt = src->pix_fmt;
+
s->seq_decode = s0->seq_decode;
s->seq_output = s0->seq_output;
s->pocTid0= s0->pocTid0;
--
2.19.1
___
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
Re: [FFmpeg-devel] [PATCH v2] ffmpeg: Add option to force a specific decode format
On 14/11/18 01:35, Fu, Linjie wrote:
>> -Original Message-
>> From: ffmpeg-devel [mailto:ffmpeg-devel-boun...@ffmpeg.org] On Behalf
>> Of Mark Thompson
>> Sent: Wednesday, November 14, 2018 09:11
>> To: ffmpeg-devel@ffmpeg.org
>> Subject: Re: [FFmpeg-devel] [PATCH v2] ffmpeg: Add option to force a
>> specific decode format
>>
>> On 14/11/18 00:50, Fu, Linjie wrote:
-Original Message-
From: ffmpeg-devel [mailto:ffmpeg-devel-boun...@ffmpeg.org] On
>> Behalf
Of Mark Thompson
Sent: Wednesday, November 14, 2018 07:44
To: ffmpeg-devel@ffmpeg.org
Subject: [FFmpeg-devel] [PATCH v2] ffmpeg: Add option to force a
>> specific
decode format
Fixes #7519.
---
doc/ffmpeg.texi | 13
fftools/ffmpeg.c | 10 ++
fftools/ffmpeg.h | 4
fftools/ffmpeg_opt.c | 47
4 files changed, 74 insertions(+)
diff --git a/doc/ffmpeg.texi b/doc/ffmpeg.texi
index 3717f22d42..d127bc0f0d 100644
--- a/doc/ffmpeg.texi
+++ b/doc/ffmpeg.texi
@@ -920,6 +920,19 @@ would be more efficient.
When doing stream copy, copy also non-key frames found at the
beginning.
+@item -decode_format[:@var{stream_specifier}]
@var{pixfmt}[,@var{pixfmt}...] (@emph{input,per-stream})
+Set the possible output formats to be used by the decoder for this
>> stream.
+If the decoder does not natively support any format in the given list for
+the input stream then decoding will fail rather than continuing with a
+different format.
+
+In general this should not be set - the decoder will select an output
+format based on the input stream parameters and available components,
and
+that will be automatically converted to whatever the output requires. It
+may be useful to force a hardware decoder supporting output in
>> multiple
+different memory types to pick a desired one, or to ensure that a
>> hardware
+decoder is used when software fallback is also available.
+
@item -init_hw_device
@var{type}[=@var{name}][:@var{device}[,@var{key=value}...]]
Initialise a new hardware device of type @var{type} called @var{name},
using the
given device parameters.
diff --git a/fftools/ffmpeg.c b/fftools/ffmpeg.c
index 38c21e944a..c651c8d3a8 100644
--- a/fftools/ffmpeg.c
+++ b/fftools/ffmpeg.c
@@ -598,6 +598,7 @@ static void ffmpeg_cleanup(int ret)
avsubtitle_free(>prev_sub.subtitle);
av_frame_free(>sub2video.frame);
av_freep(>filters);
+av_freep(>decode_formats);
av_freep(>hwaccel_device);
av_freep(>dts_buffer);
@@ -2800,6 +2801,15 @@ static enum AVPixelFormat
get_format(AVCodecContext *s, const enum AVPixelFormat
const AVCodecHWConfig *config = NULL;
int i;
+if (ist->decode_formats) {
+for (i = 0; ist->decode_formats[i] != AV_PIX_FMT_NONE; i++) {
+if (ist->decode_formats[i] == *p)
+break;
+}
+if (ist->decode_formats[i] != *p)
+continue;
+}
+
if (!(desc->flags & AV_PIX_FMT_FLAG_HWACCEL))
break;
diff --git a/fftools/ffmpeg.h b/fftools/ffmpeg.h
index eb1eaf6363..b06fd18b1c 100644
--- a/fftools/ffmpeg.h
+++ b/fftools/ffmpeg.h
@@ -125,6 +125,8 @@ typedef struct OptionsContext {
intnb_ts_scale;
SpecifierOpt *dump_attachment;
intnb_dump_attachment;
+SpecifierOpt *decode_formats;
+intnb_decode_formats;
SpecifierOpt *hwaccels;
intnb_hwaccels;
SpecifierOpt *hwaccel_devices;
@@ -334,6 +336,8 @@ typedef struct InputStream {
int top_field_first;
int guess_layout_max;
+enum AVPixelFormat *decode_formats;
+
int autorotate;
int fix_sub_duration;
diff --git a/fftools/ffmpeg_opt.c b/fftools/ffmpeg_opt.c
index d4851a2cd8..63bb05053b 100644
--- a/fftools/ffmpeg_opt.c
+++ b/fftools/ffmpeg_opt.c
@@ -701,6 +701,7 @@ static void add_input_streams(OptionsContext *o,
AVFormatContext *ic)
AVStream *st = ic->streams[i];
AVCodecParameters *par = st->codecpar;
InputStream *ist = av_mallocz(sizeof(*ist));
+char *decode_formats = NULL;
char *framerate = NULL, *hwaccel_device = NULL;
const char *hwaccel = NULL;
char *hwaccel_output_format = NULL;
@@ -797,6 +798,49 @@ static void add_input_streams(OptionsContext
>> *o,
AVFormatContext *ic)
ist->top_field_first = -1;
[FFmpeg-devel] [PATCH v2] hwcontext_vaapi: Add option to set driver name
For example: -init_hw_device vaapi:/dev/dri/renderD128,driver=foo
This may be more convenient that using the environment variable, and allows
loading different drivers for different devices in the same process.
---
On 14/11/18 12:05, Michael Niedermayer wrote:
> On Wed, Nov 14, 2018 at 12:21:07AM +, Mark Thompson wrote:
>> ...
>> +vas = vaSetDriverName(display, driver->value);
>
> this breaks build here on ubuntu:
>
> CClibavutil/hwcontext_vaapi.o
> libavutil/hwcontext_vaapi.c: In function ‘vaapi_device_create’:
> libavutil/hwcontext_vaapi.c:1537:9: error: implicit declaration of function
> ‘vaSetDriverName’ [-Werror=implicit-function-declaration]
> vas = vaSetDriverName(display, driver->value);
> ^
> cc1: some warnings being treated as errors
> make: *** [libavutil/hwcontext_vaapi.o] Error 1
> make: Target `all' not remade because of errors.
Ah, apologies - that API is newer than I thought.
Here's a new version with suitable guards.
Thanks,
- Mark
libavutil/hwcontext_vaapi.c | 20
1 file changed, 20 insertions(+)
diff --git a/libavutil/hwcontext_vaapi.c b/libavutil/hwcontext_vaapi.c
index 8624369bb9..4a4ecc44d7 100644
--- a/libavutil/hwcontext_vaapi.c
+++ b/libavutil/hwcontext_vaapi.c
@@ -1469,6 +1469,7 @@ static int vaapi_device_create(AVHWDeviceContext *ctx,
const char *device,
{
VAAPIDevicePriv *priv;
VADisplay display = NULL;
+const AVDictionaryEntry *driver;
priv = av_mallocz(sizeof(*priv));
if (!priv)
@@ -1530,6 +1531,25 @@ static int vaapi_device_create(AVHWDeviceContext *ctx,
const char *device,
return AVERROR(EINVAL);
}
+driver = av_dict_get(opts, "driver", NULL, 0);
+if (driver) {
+#if VA_CHECK_VERSION(0, 38, 0)
+VAStatus vas;
+vas = vaSetDriverName(display, driver->value);
+if (vas != VA_STATUS_SUCCESS) {
+av_log(ctx, AV_LOG_ERROR, "Failed to set driver name to "
+ "%s: %d (%s).\n", driver->value, vas, vaErrorStr(vas));
+vaTerminate(display);
+return AVERROR_UNKNOWN;
+}
+#else
+av_log(ctx, AV_LOG_WARNING, "Driver name setting is not "
+ "supported with this VAAPI version.\n");
+vaTerminate(display);
+return AVERROR(ENOSYS);
+#endif
+}
+
return vaapi_device_connect(ctx, display);
}
--
2.19.1
___
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
[FFmpeg-devel] [PATCH] avcodec/libvpxenc: add VP8 temporal scalability configuration options
This commit adds configuration options to libvpxenc.c that can be used to
enable VP8 temporal scalability. It also adds a way to programmatically set the
per-frame encoding flags which can be used to control usage and updates of
reference frames while encoding with temporal scalability enabled.
---
doc/encoders.texi | 28 +++
libavcodec/libvpxenc.c | 81 ++
2 files changed, 109 insertions(+)
diff --git a/doc/encoders.texi b/doc/encoders.texi
index c9464ca7b3..39c3830dd5 100644
--- a/doc/encoders.texi
+++ b/doc/encoders.texi
@@ -1655,6 +1655,34 @@ Set number of frames to look ahead for frametype and
ratecontrol.
@item error-resilient
Enable error resiliency features.
+@item VP8-specific options
+@table @option
+@item ts-parameters
+Sets the temporal scalability configuration using a :-separated list of
+key=value pairs. For example, to specify temporal scalability parameters
+with @code{ffmpeg}:
+@example
+ffmpeg -i INPUT -c:v libvpx -ts-parameters ts_number_layers=3:\
+ts_target_bitrate=25,50,100:ts_rate_decimator=4,2,1:\
+ts_periodicity=4:ts_layer_id=0,2,1,2 OUTPUT
+@end example
+Below is a brief explanation of each of the parameters, please
+refer to @code{struct vpx_codec_enc_cfg} in @code{vpx/vpx_encoder.h} for more
+details.
+@table @option
+@item ts_number_layers
+Number of temporal coding layers.
+@item ts_target_bitrate
+Target bitrate for each temporal layer.
+@item ts_rate_decimator
+Frame rate decimation factor for each temporal layer.
+@item ts_periodicity
+Length of the sequence defining frame temporal layer membership.
+@item ts_layer_id
+Template defining the membership of frames to temporal layers.
+@end table
+@end table
+
@item VP9-specific options
@table @option
@item lossless
diff --git a/libavcodec/libvpxenc.c b/libavcodec/libvpxenc.c
index ad440a9c21..774fda457e 100644
--- a/libavcodec/libvpxenc.c
+++ b/libavcodec/libvpxenc.c
@@ -33,6 +33,7 @@
#include "libavutil/avassert.h"
#include "libvpx.h"
#include "profiles.h"
+#include "libavutil/avstring.h"
#include "libavutil/base64.h"
#include "libavutil/common.h"
#include "libavutil/internal.h"
@@ -98,6 +99,8 @@ typedef struct VPxEncoderContext {
int rc_undershoot_pct;
int rc_overshoot_pct;
+char *vp8_ts_parameters;
+
// VP9-only
int lossless;
int tile_columns;
@@ -169,6 +172,7 @@ static av_cold void dump_enc_cfg(AVCodecContext *avctx,
{
int width = -30;
int level = AV_LOG_DEBUG;
+int i;
av_log(avctx, level, "vpx_codec_enc_cfg\n");
av_log(avctx, level, "generic settings\n"
@@ -208,6 +212,25 @@ static av_cold void dump_enc_cfg(AVCodecContext *avctx,
" %*s%u\n %*s%u\n",
width, "rc_undershoot_pct:", cfg->rc_undershoot_pct,
width, "rc_overshoot_pct:", cfg->rc_overshoot_pct);
+av_log(avctx, level, "temporal layering settings\n"
+ " %*s%u\n", width, "ts_number_layers:", cfg->ts_number_layers);
+av_log(avctx, level,
+ "\n %*s", width, "ts_target_bitrate:");
+for (i = 0; i < VPX_TS_MAX_LAYERS; i++)
+av_log(avctx, level, "%u ", cfg->ts_target_bitrate[i]);
+av_log(avctx, level, "\n");
+av_log(avctx, level,
+ "\n %*s", width, "ts_rate_decimator:");
+for (i = 0; i < VPX_TS_MAX_LAYERS; i++)
+av_log(avctx, level, "%u ", cfg->ts_rate_decimator[i]);
+av_log(avctx, level, "\n");
+av_log(avctx, level,
+ "\n %*s%u\n", width, "ts_periodicity:", cfg->ts_periodicity);
+av_log(avctx, level,
+ "\n %*s", width, "ts_layer_id:");
+for (i = 0; i < VPX_TS_MAX_PERIODICITY; i++)
+av_log(avctx, level, "%u ", cfg->ts_layer_id[i]);
+av_log(avctx, level, "\n");
av_log(avctx, level, "decoder buffer model\n"
" %*s%u\n %*s%u\n %*s%u\n",
width, "rc_buf_sz:", cfg->rc_buf_sz,
@@ -325,6 +348,40 @@ static av_cold int vpx_free(AVCodecContext *avctx)
return 0;
}
+static void vp8_ts_parse_int_array(int *dest, char *value, size_t value_len,
int max_entries)
+{
+int dest_idx = 0;
+char *saveptr = NULL;
+char *token = av_strtok(value, ",", );
+
+while (token && dest_idx < max_entries)
+{
+dest[dest_idx++] = strtoul(token, NULL, 10);
+token = av_strtok(NULL, ",", );
+}
+}
+
+static int vp8_ts_param_parse(struct vpx_codec_enc_cfg *enccfg, char *key,
char *value)
+{
+size_t value_len = strlen(value);
+
+if (!value_len)
+return -1;
+
+if (!strcmp(key, "ts_number_layers"))
+enccfg->ts_number_layers = strtoul(value, , 10);
+else if (!strcmp(key, "ts_target_bitrate"))
+vp8_ts_parse_int_array(enccfg->ts_target_bitrate, value, value_len,
VPX_TS_MAX_LAYERS);
+else if (!strcmp(key, "ts_rate_decimator"))
+ vp8_ts_parse_int_array(enccfg->ts_rate_decimator, value, value_len,
VPX_TS_MAX_LAYERS);
+else if (!strcmp(key, "ts_periodicity"))
+
Re: [FFmpeg-devel] [PATCH] avcodec/vc1: correct aspect ratio calculation
Hi
On Wed, Nov 14, 2018 at 10:57:25AM +0100, Jerome Borsboom wrote:
> According to VC-1 spec:
> * Display size defaults to max coded size when not explicitly set in
> sequence header
> * Aspect ratio in the sequence header refers to the Display size elements.
> Therefore, the aspect ratio for the coded samples (SAR) needs to take into
> account the scaling from coded size to display size, and the aspect ratio
> of the display size elements.
>
> Signed-off-by: Jerome Borsboom
> ---
> VC-1 spec assumes that the output of the decoder, i.e. the pixel matrix with
> dimensions coded_width x coded_height, is scaled to display size, i.e. a pixel
> matrix of display_horiz_size x display_vert_size. This may introduce part of
> the sample aspect ratio when the sizes of the two pixel matrices are not
> equal.
> A further part of the sample aspect ratio is optionally specified in the
> sequence
> header as the aspect ratio of the display size pixels. This patch takes both
> aspect ratios into account and aims to be correct even when the coded image
> includes overscan regions.
>
> libavcodec/vc1.c | 38 +-
> libavcodec/vc1.h | 2 ++
> 2 files changed, 23 insertions(+), 17 deletions(-)
>
> diff --git a/libavcodec/vc1.c b/libavcodec/vc1.c
> index 3581d87b57..efc6edc4b0 100644
> --- a/libavcodec/vc1.c
> +++ b/libavcodec/vc1.c
> @@ -442,30 +442,24 @@ static int decode_sequence_header_adv(VC1Context *v,
> GetBitContext *gb)
> }
> v->s.max_b_frames = v->s.avctx->max_b_frames = 7;
> if (get_bits1(gb)) { //Display Info - decoding is not affected by it
> -int w, h, ar = 0;
> +int ar = 0;
> av_log(v->s.avctx, AV_LOG_DEBUG, "Display extended info:\n");
> -w = get_bits(gb, 14) + 1;
> -h = get_bits(gb, 14) + 1;
> -av_log(v->s.avctx, AV_LOG_DEBUG, "Display dimensions: %ix%i\n", w,
> h);
> +v->disp_horiz_size = get_bits(gb, 14) + 1;
> +v->disp_vert_size = get_bits(gb, 14) + 1;
> +av_log(v->s.avctx, AV_LOG_DEBUG, "Display dimensions: %ix%i\n",
> + v->disp_horiz_size, v->disp_vert_size);
> if (get_bits1(gb))
> ar = get_bits(gb, 4);
> if (ar && ar < 14) {
> -v->s.avctx->sample_aspect_ratio = ff_vc1_pixel_aspect[ar];
> +v->aspect_ratio = ff_vc1_pixel_aspect[ar];
> } else if (ar == 15) {
> -w = get_bits(gb, 8) + 1;
> -h = get_bits(gb, 8) + 1;
> -v->s.avctx->sample_aspect_ratio = (AVRational){w, h};
> +v->aspect_ratio = (AVRational){get_bits(gb, 8) + 1, get_bits(gb,
> 8) + 1};
iam not sure this is valid C and not undefined
but either way this patch breaks fate
TESTvc1-ism
--- ./tests/ref/fate/vc1-ism2018-11-13 19:52:23.489023763 +0100
+++ tests/data/fate/vc1-ism 2018-11-14 21:50:11.522992878 +0100
@@ -2,7 +2,7 @@
#media_type 0: video
#codec_id 0: rawvideo
#dimensions 0: 240x104
-#sar 0: 156/156
+#sar 0: 13/30
0, 0, 0,1,37440, 0xd1bc5235
0, 2, 2,1,37440, 0x158e6167
0, 3, 3,1,37440, 0x0faa4481
Test vc1-ism failed. Look at tests/data/fate/vc1-ism.err for details.
make: *** [fate-vc1-ism] Error 1
[...]
--
Michael GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB
it is not once nor twice but times without number that the same ideas make
their appearance in the world. -- Aristotle
signature.asc
Description: PGP signature
___
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
[FFmpeg-devel] [PATCH 1/1] avcodec/vaapi_encode: add frame-skip func
frame-skip is required to implement network
bandwidth self-adaptive vaapi encoding.
To make a frame skipped, allocate its frame
side data of AV_FRAME_DATA_SKIP_FRAME type
and set its value to 1.
Signed-off-by: Jing SUN
---
libavcodec/vaapi_encode.c | 132 --
libavcodec/vaapi_encode.h | 5 ++
libavutil/frame.c | 1 +
libavutil/frame.h | 5 ++
4 files changed, 139 insertions(+), 4 deletions(-)
diff --git a/libavcodec/vaapi_encode.c b/libavcodec/vaapi_encode.c
index 2fe8501..ea06404 100644
--- a/libavcodec/vaapi_encode.c
+++ b/libavcodec/vaapi_encode.c
@@ -23,6 +23,7 @@
#include "libavutil/common.h"
#include "libavutil/log.h"
#include "libavutil/pixdesc.h"
+#include "libavutil/intreadwrite.h"
#include "vaapi_encode.h"
#include "avcodec.h"
@@ -103,6 +104,47 @@ static int vaapi_encode_make_param_buffer(AVCodecContext
*avctx,
return 0;
}
+static int vaapi_encode_check_if_skip(AVCodecContext *avctx,
+ VAAPIEncodePicture *pic)
+{
+AVFrameSideData *fside = NULL;
+VAAPIEncodeContext *ctx = avctx->priv_data;
+VAAPIEncodePicture *cur = NULL;
+int i = 0;
+
+if (!pic || !pic->input_image)
+return AVERROR(EINVAL);
+
+fside = av_frame_get_side_data(pic->input_image, AV_FRAME_DATA_SKIP_FRAME);
+if (fside)
+pic->skipped_flag = AV_RL8(fside->data);
+else
+pic->skipped_flag = 0;
+
+if (0 == pic->skipped_flag)
+return 0;
+
+if ((pic->type == PICTURE_TYPE_IDR) || (pic->type == PICTURE_TYPE_I)) {
+av_log(avctx, AV_LOG_INFO, "Can't skip IDR/I pic
%"PRId64"/%"PRId64".\n",
+ pic->display_order, pic->encode_order);
+pic->skipped_flag = 0;
+return 0;
+}
+
+for (cur = ctx->pic_start; cur; cur = cur->next) {
+for (i=0; i < cur->nb_refs; ++i) {
+if (cur->refs[i] == pic) {
+av_log(avctx, AV_LOG_INFO, "Can't skip ref pic
%"PRId64"/%"PRId64".\n",
+ pic->display_order, pic->encode_order);
+pic->skipped_flag = 0;
+return 0;
+}
+}
+}
+
+return 0;
+}
+
static int vaapi_encode_wait(AVCodecContext *avctx,
VAAPIEncodePicture *pic)
{
@@ -418,6 +460,59 @@ static int vaapi_encode_issue(AVCodecContext *avctx,
}
}
+err = vaapi_encode_check_if_skip(avctx, pic);
+if (err != 0)
+av_log(avctx, AV_LOG_ERROR, "Fail to check if skip.\n");
+
+if (pic->skipped_flag) {
+av_log(avctx, AV_LOG_INFO, "Skip pic %"PRId64"/%"PRId64" as
requested.\n",
+ pic->display_order, pic->encode_order);
+
+++ctx->skipped_pic_count;
+pic->encode_issued = 1;
+
+return 0;
+} else if (ctx->skipped_pic_count > 0) {
+VABufferID skip_param_id;
+VAEncMiscParameterBuffer *misc_param;
+VAEncMiscParameterSkipFrame *skip_param;
+
+err = vaapi_encode_make_param_buffer(avctx, pic,
+ VAEncMiscParameterBufferType, NULL,
+ (sizeof(VAEncMiscParameterBuffer) +
+ sizeof(VAEncMiscParameterSkipFrame)));
+if (err < 0)
+goto fail;
+
+skip_param_id = pic->param_buffers[pic->nb_param_buffers-1];
+
+vas = vaMapBuffer(ctx->hwctx->display,
+ skip_param_id,
+ (void **)_param);
+if (vas != VA_STATUS_SUCCESS) {
+av_log(avctx, AV_LOG_ERROR, "Failed to map skip-frame buffer: "
+ "%d (%s).\n", vas, vaErrorStr(vas));
+err = AVERROR(EIO);
+goto fail;
+}
+
+misc_param->type =
(VAEncMiscParameterType)VAEncMiscParameterTypeSkipFrame;
+skip_param = (VAEncMiscParameterSkipFrame *)misc_param->data;
+skip_param->skip_frame_flag = 1;
+skip_param->num_skip_frames = ctx->skipped_pic_count;
+skip_param->size_skip_frames = 0;
+
+vas = vaUnmapBuffer(ctx->hwctx->display, skip_param_id);
+if (vas != VA_STATUS_SUCCESS) {
+av_log(avctx, AV_LOG_ERROR, "Failed to unmap skip-frame buffer: "
+ "%d (%s).\n", vas, vaErrorStr(vas));
+err = AVERROR(EIO);
+goto fail;
+}
+
+ctx->skipped_pic_count = 0;
+}
+
vas = vaBeginPicture(ctx->hwctx->display, ctx->va_context,
pic->input_surface);
if (vas != VA_STATUS_SUCCESS) {
@@ -500,9 +595,28 @@ static int vaapi_encode_output(AVCodecContext *avctx,
VAStatus vas;
int err;
-err = vaapi_encode_wait(avctx, pic);
-if (err < 0)
-return err;
+if (!pic->skipped_flag) {
+err = vaapi_encode_wait(avctx, pic);
+if (err < 0)
+return err;
+} else {
+av_frame_free(>input_image);
+pic->encode_complete = 1;
+
+err = av_new_packet(pkt, 0);
+if (err < 0)
+
Re: [FFmpeg-devel] [PATCH v2] lavf/isom: add Dolby Vision sample entry codes for HEVC and H.264
On Sat, Nov 10, 2018 at 3:36 PM Carl Eugen Hoyos wrote: > > 2018-11-08 17:34 GMT+01:00, Jan Ekström : > > > Then regarding the rest. Do I understand correctly that you do > > not trust dolby's specification and/or mpeg-4 ra and would only > > like identifiers be added that have been seen in the wild? > > This is not trust-related, I simply believe this patch is good > example for why patches should be tested with samples. > > Carl Eugen So far: - 2/4 identifiers have been proven to match the darn specification regarding the sample identifier and the fact that the standard decoder initialization box is in there. I dislike Dolby as much as the other person, but this is so far a good track record and given the wording of the specification I would tend to trust it. - VLC has added these identifiers (http://git.videolan.org/?p=vlc.git;a=commit;h=dbb0f6c7353aad4cbbb14d88f9409c8419c26bcd) - Chromium has added these identifiers (https://codereview.chromium.org/2640113004/ - it goes to a separate codec on their side since they just want to pass it to the specific decoder interfaces that can handle the profiles fully instead of actually decoding them themselves). So what if I just make a patch without dvh1 that you have already implemented, and then we call that a day? Jan ___ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
Re: [FFmpeg-devel] [PATCH 1/2] lavu: add locale-independent sscanf implementation
On 11/12/18, Paul B Mahol wrote: > Copied and adopted from musl implementation. > * converted all 'long double' to 'double' > * removed %m support > > Signed-off-by: Paul B Mahol > --- > libavutil/Makefile | 1 + > libavutil/avsscanf.c | 971 +++ > libavutil/avstring.c | 1 + > libavutil/avstring.h | 6 + > 4 files changed, 979 insertions(+) > create mode 100644 libavutil/avsscanf.c If there are no strong opinions against this patch set i will apply it and work on converting more code to av_sscanf(). ___ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
Re: [FFmpeg-devel] [PATCH] lavf/dashenc: set FLAC manifest codec string to "flac"
On Wed, Nov 14, 2018 at 11:55 AM Paul B Mahol wrote: > > On 11/14/18, Jan Ekström wrote: > > Internally in ISOBMFF the FLAC-in-ISOBMFF draft uses the "fLaC" > > as the identifier for FLACSampleEntry, but what the browsers > > seem to eat for MPEG-DASH manifests is plain "flac". > > > > Verified by auri_ on IRC to play with the major browsers. > > --- > > libavformat/dashenc.c | 1 + > > 1 file changed, 1 insertion(+) > > > > lgtm Thanks. Applied with an improved commit message that I think explains the situation with regards to unavailability of a specification for this value utilized in the MPEG-DASH manifest. Jan ___ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
Re: [FFmpeg-devel] [PATCH] avcodec/vc1: correct aspect ratio calculation
2018-11-14 17:16 GMT+01:00, Jerome Borsboom : >>> +av_reduce(>sample_aspect_ratio.num, >>> + >sample_aspect_ratio.den, >>> + v->disp_horiz_size * v->aspect_ratio.num * h, >>> + v->disp_vert_size * v->aspect_ratio.den * w, >>> + 1 << 30); >> >>> +ff_set_sar(avctx, avctx->sample_aspect_ratio); >> >> I would have expected these two statements to be redundant - >> am I wrong? > av_reduce takes out the greatest common divisor and puts a bound on the > individual fraction numbers. ff_set_sar checks that the SAR makes sense > in relation to the image size. Although, the code is reassigning > avctx->sample_aspect_ratio to itself in ff_set_sar, the different checks > make both statements necessary. Thank you! Carl Eugen ___ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
[FFmpeg-devel] [PATCH] web/index: add news entry for release 4.1
Signed-off-by: James Almer --- src/index | 50 ++ 1 file changed, 50 insertions(+) diff --git a/src/index b/src/index index bbfca29..0dcf6c1 100644 --- a/src/index +++ b/src/index @@ -37,6 +37,56 @@ News + November 6th, 2018, FFmpeg 4.1 "al-Khwarizmi" + +FFmpeg 4.1 "al-Khwarizmi", a new +major release, is now available! Some of the highlights: + + +deblock filter +tmix filter +amplify filter +fftdnoiz filter +aderivative and aintegral audio filters +pal75bars and pal100bars video filter sources +mbedTLS based TLS support +adeclick and adeclip filters +libtensorflow backend for DNN based filters like srcnn +VC1 decoder is now bit-exact +ATRAC9 decoder +lensfun wrapper filter +colorconstancy filter +AVS2 video decoder via libdavs2 +IMM4 video decoder +Brooktree ProSumer video decoder +MatchWare Screen Capture Codec decoder +WinCam Motion Video decoder +1D LUT filter (lut1d) +RemotelyAnywhere Screen Capture decoder +cue and acue filters +Support for AV1 in MP4 and Matroska/WebM +transpose_npp filter +AVS2 video encoder via libxavs2 +amultiply filter +Block-Matching 3d (bm3d) denoising filter +acrossover filter +ilbc decoder +audio denoiser as afftdn filter +AV1 parser +sinc audio filter source +chromahold filter +setparams filter +vibrance filter +S12M timecode decoding in h264 +xstack filter +(a)graphmonitor filter +yadif_cuda filter + + +We strongly recommend users, distributors, and system integrators to +upgrade unless they use current git master. + + April 20th, 2018, FFmpeg 4.0 "Wu" FFmpeg 4.0 "Wu", a new -- 2.19.1 ___ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
Re: [FFmpeg-devel] [PATCH] avcodec/vc1: correct aspect ratio calculation
>> +av_reduce(>sample_aspect_ratio.num, >> + >sample_aspect_ratio.den, >> + v->disp_horiz_size * v->aspect_ratio.num * h, >> + v->disp_vert_size * v->aspect_ratio.den * w, >> + 1 << 30); > >> +ff_set_sar(avctx, avctx->sample_aspect_ratio); > > I would have expected these two statements to be redundant - > am I wrong? > > Carl Eugen I am not fully sure that I understand your comment. av_reduce takes out the greatest common divisor and puts a bound on the individual fraction numbers. ff_set_sar checks that the SAR makes sense in relation to the image size. Although, the code is reassigning avctx->sample_aspect_ratio to itself in ff_set_sar, the different checks make both statements necessary. In addition, the large values that results from 'v->disp_horiz_size * v->aspect_ratio.num * h' and 'v->disp_vert_size * v->aspect_ratio.den * w', can in most cases be efficiently reduced. These two statements are moved to the entrypoint parser, because the coded size is an entrypoint element. At the position where the av_reduce statement was previously located, the v->s.avctx->height and v->s.avctx->width could contain values from the previous entrypoint header or be zero when parsing the first sequence header. Regards, Jerome ___ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
Re: [FFmpeg-devel] [PATCH] lavf/dashenc: set FLAC manifest codec string to "flac"
On Wed, Nov 14, 2018, 17:22 Carl Eugen Hoyos 2018-11-14 2:21 GMT+01:00, Jan Ekström : > > Internally in ISOBMFF the FLAC-in-ISOBMFF draft uses the "fLaC" > > as the identifier for FLACSampleEntry, but what the browsers > > seem to eat for MPEG-DASH manifests is plain "flac". > > Does this patch also answer the recent discussion about > specifications vs real-world tests? > No. The ISOBMFF side is followed properly, and nobody cared to test flac until now with dash manifests. The default code path utilizes the ISOBMFF identifier for dash manifest entries, which matches things such as avc/hevc/aac etc. Not to mention that there doesn't seem to be a spec for the dash manifest side of things for flac. At all. It just seems to follow the mime type. So no, this is not the same thing. At all. Jan > ___ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
Re: [FFmpeg-devel] [FFmpeg-cvslog] avcodec/libdav1d: fix build after a recent API break
On 11/14/2018 12:38 PM, Nicolas George wrote:
> James Almer (2018-11-14):
>> -res = dav1d_decode(dav1d->c, data, );
>> +res = dav1d_send_data(dav1d->c, data);
>> +if (res < 0) {
>> +if (res == -EINVAL)
>> +res = AVERROR_INVALIDDATA;
>> +if (res != -EAGAIN)
>> +return res;
>> +}
>> +
>> +res = dav1d_get_picture(dav1d->c, );
>
> I just noticed this: the lack of AVERROR() looks very suspicious. It
> seems this library uses directly -errno values, which will coincide with
> AVERROR() on most systems but not all.
>
> Regards,
I considered that, but I don't think libdav1d is supported on systems
where errno values are already negated, as every single one of their
internal error checks would fail. So it's probably safe to do this.
___
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
Re: [FFmpeg-devel] [FFmpeg-cvslog] avcodec/libdav1d: fix build after a recent API break
James Almer (2018-11-14):
> -res = dav1d_decode(dav1d->c, data, );
> +res = dav1d_send_data(dav1d->c, data);
> +if (res < 0) {
> +if (res == -EINVAL)
> +res = AVERROR_INVALIDDATA;
> +if (res != -EAGAIN)
> +return res;
> +}
> +
> +res = dav1d_get_picture(dav1d->c, );
I just noticed this: the lack of AVERROR() looks very suspicious. It
seems this library uses directly -errno values, which will coincide with
AVERROR() on most systems but not all.
Regards,
--
Nicolas George
signature.asc
Description: Digital signature
___
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
Re: [FFmpeg-devel] [FFmpeg-cvslog] avcodec/libdav1d: fix build after a recent API break
2018-11-14 16:19 GMT+01:00, James Almer :
> On 11/14/2018 12:03 PM, Carl Eugen Hoyos wrote:
>> 2018-11-14 14:06 GMT+01:00, James Almer :
>>> ffmpeg | branch: master | James Almer | Wed Nov 14
>>> 10:05:25 2018 -0300| [752659327d4ac73640781376d214a26765f971f4] |
>>> committer:
>>> James Almer
>>>
>>> avcodec/libdav1d: fix build after a recent API break
>>>
>>> Signed-off-by: James Almer
>>>
http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=752659327d4ac73640781376d214a26765f971f4
>>> ---
>>>
>>> libavcodec/libdav1d.c | 12 +---
>>> 1 file changed, 9 insertions(+), 3 deletions(-)
>>>
>>> diff --git a/libavcodec/libdav1d.c b/libavcodec/libdav1d.c
>>> index 873adfda40..b19aee4e9e 100644
>>> --- a/libavcodec/libdav1d.c
>>> +++ b/libavcodec/libdav1d.c
>>> @@ -140,12 +140,18 @@ static int libdav1d_receive_frame(AVCodecContext
>>> *c,
>>> AVFrame *frame)
>>> }
>>>
>>> av_fifo_generic_write(dav1d->cache, , sizeof(pkt),
>>> libdav1d_fifo_write);
>>> -} else {
>>> -data = NULL;
>>> }
>>> }
>>>
>>> -res = dav1d_decode(dav1d->c, data, );
>>> +res = dav1d_send_data(dav1d->c, data);
>>> +if (res < 0) {
>>> +if (res == -EINVAL)
>>> +res = AVERROR_INVALIDDATA;
>>
>>> +if (res != -EAGAIN)
>>
>> How does this line make sense?
>
> We're meant to keep going and fetch decoded frames even if the library
> refuses new data. Eventually it will consume it once its internal data
> buffer is depleted.
>
> We're only mean to abort on other errno values, like -EINVAL (Which for
> ffmpeg is AVERROR_INVALIDDATA).
Thank you!
Carl Eugen
___
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
Re: [FFmpeg-devel] [PATCH] lavf/dashenc: set FLAC manifest codec string to "flac"
2018-11-14 2:21 GMT+01:00, Jan Ekström : > Internally in ISOBMFF the FLAC-in-ISOBMFF draft uses the "fLaC" > as the identifier for FLACSampleEntry, but what the browsers > seem to eat for MPEG-DASH manifests is plain "flac". Does this patch also answer the recent discussion about specifications vs real-world tests? Carl Eugen ___ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
Re: [FFmpeg-devel] [FFmpeg-cvslog] avcodec/libdav1d: fix build after a recent API break
On 11/14/2018 12:03 PM, Carl Eugen Hoyos wrote:
> 2018-11-14 14:06 GMT+01:00, James Almer :
>> ffmpeg | branch: master | James Almer | Wed Nov 14
>> 10:05:25 2018 -0300| [752659327d4ac73640781376d214a26765f971f4] | committer:
>> James Almer
>>
>> avcodec/libdav1d: fix build after a recent API break
>>
>> Signed-off-by: James Almer
>>
>>> http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=752659327d4ac73640781376d214a26765f971f4
>> ---
>>
>> libavcodec/libdav1d.c | 12 +---
>> 1 file changed, 9 insertions(+), 3 deletions(-)
>>
>> diff --git a/libavcodec/libdav1d.c b/libavcodec/libdav1d.c
>> index 873adfda40..b19aee4e9e 100644
>> --- a/libavcodec/libdav1d.c
>> +++ b/libavcodec/libdav1d.c
>> @@ -140,12 +140,18 @@ static int libdav1d_receive_frame(AVCodecContext *c,
>> AVFrame *frame)
>> }
>>
>> av_fifo_generic_write(dav1d->cache, , sizeof(pkt),
>> libdav1d_fifo_write);
>> -} else {
>> -data = NULL;
>> }
>> }
>>
>> -res = dav1d_decode(dav1d->c, data, );
>> +res = dav1d_send_data(dav1d->c, data);
>> +if (res < 0) {
>> +if (res == -EINVAL)
>> +res = AVERROR_INVALIDDATA;
>
>> +if (res != -EAGAIN)
>
> How does this line make sense?
We're meant to keep going and fetch decoded frames even if the library
refuses new data. Eventually it will consume it once its internal data
buffer is depleted.
We're only mean to abort on other errno values, like -EINVAL (Which for
ffmpeg is AVERROR_INVALIDDATA).
>
>> +return res;
>
> Carl Eugen
> ___
> ffmpeg-devel mailing list
> ffmpeg-devel@ffmpeg.org
> http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
>
___
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
Re: [FFmpeg-devel] [PATCH] avcodec/vc1: correct aspect ratio calculation
2018-11-14 10:57 GMT+01:00, Jerome Borsboom : > +av_reduce(>sample_aspect_ratio.num, > + >sample_aspect_ratio.den, > + v->disp_horiz_size * v->aspect_ratio.num * h, > + v->disp_vert_size * v->aspect_ratio.den * w, > + 1 << 30); > +ff_set_sar(avctx, avctx->sample_aspect_ratio); I would have expected these two statements to be redundant - am I wrong? Carl Eugen ___ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
Re: [FFmpeg-devel] [FFmpeg-cvslog] avcodec/libdav1d: fix build after a recent API break
2018-11-14 14:06 GMT+01:00, James Almer :
> ffmpeg | branch: master | James Almer | Wed Nov 14
> 10:05:25 2018 -0300| [752659327d4ac73640781376d214a26765f971f4] | committer:
> James Almer
>
> avcodec/libdav1d: fix build after a recent API break
>
> Signed-off-by: James Almer
>
>> http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=752659327d4ac73640781376d214a26765f971f4
> ---
>
> libavcodec/libdav1d.c | 12 +---
> 1 file changed, 9 insertions(+), 3 deletions(-)
>
> diff --git a/libavcodec/libdav1d.c b/libavcodec/libdav1d.c
> index 873adfda40..b19aee4e9e 100644
> --- a/libavcodec/libdav1d.c
> +++ b/libavcodec/libdav1d.c
> @@ -140,12 +140,18 @@ static int libdav1d_receive_frame(AVCodecContext *c,
> AVFrame *frame)
> }
>
> av_fifo_generic_write(dav1d->cache, , sizeof(pkt),
> libdav1d_fifo_write);
> -} else {
> -data = NULL;
> }
> }
>
> -res = dav1d_decode(dav1d->c, data, );
> +res = dav1d_send_data(dav1d->c, data);
> +if (res < 0) {
> +if (res == -EINVAL)
> +res = AVERROR_INVALIDDATA;
> +if (res != -EAGAIN)
How does this line make sense?
> +return res;
Carl Eugen
___
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
Re: [FFmpeg-devel] [PATCH] lavf/dashenc: Add support for per-stream container type selection.
On 11/14/18 4:52 PM, Jeyapal, Karthick wrote: On 11/14/18 1:41 PM, Andrey Semashev wrote: On 11/12/18 3:55 PM, Andrey Semashev wrote: Suppose you want to create a DASH stream consisting of VP8, H264, Vorbis and Opus substreams to cover the best compatibility with clients (which are mostly web browsers, but not necessarilly all of them). AFAIK, you cannot put all these codecs neither in mp4 nor webm. An "auto" option would put Opus in webm container, leaving clients not supporting webm not able to play audio. With explicit container selection one could put Opus content in mp4 or both webm and mp4 (in which case the substream would be duplicated). An example of a client that is picky about container format is Safari on OS X High Sierra. I don't have one to test, but reportedly it supports Opus only in caf format, and I've read someone hacked it to play Opus in mp4 if disguised as AAC. I know lavf/dashenc doesn't support caf (yet) but it may support it in the future, so the container format selection would become even more relevant then. So, what do we decide about this patch? I understand your point. But all your examples are still theoretical(say caf is not supported in dashenc. And opus example doesn't mention any sample clients for that use-case). My view is that, if we can find one practical example where this advanced patch will be of immediate use(which cannot be done by auto option), then let us review and push this patch. If we can't find any immediate practical use-case, which could be tested and verified in popular clients then let's go for auto option. Ok, thanks for your comments. ___ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
Re: [FFmpeg-devel] [PATCH] lavf/dashenc: Add support for per-stream container type selection.
On 11/14/18 1:41 PM, Andrey Semashev wrote: > On 11/12/18 3:55 PM, Andrey Semashev wrote: > > On 11/12/18 3:12 PM, Jeyapal, Karthick wrote: > >> > >> On 11/12/18 5:20 PM, Andrey Semashev wrote: > >>> On 11/12/18 8:20 AM, Jeyapal, Karthick wrote: > > On 11/8/18 10:27 PM, Andrey Semashev wrote: > > This commit restores the ability to create DASH streams with codecs > > that require different containers that was lost after commit > > 2efdbf7367989cf9d296c25fa3d2aff8d6e25fdd. It extends the > > dash_segment_type > > option syntax to allow to specify segment container types for > > individual > > streams, in addition to the default container type that is applied to > > all streams. The extended syntax is backward compatible with the > > previous > > syntax. > Thanks for sending the patch. I understand your requirement completely. > But I feel that this option for mapping streams with container > format is little confusing. Also, the relevant code is relatively > big, and thus difficult to maintain in future. > I have a middle ground suggestion. If your goal is to achieve the > earlier behavior broken commits, then I propose the following. > Option "dash_segment_type" could take one more option "auto" > (instead of mp4 or webm). > When "auto" is chosen, the muxer could choose webm format for VP8, > VP9, vorbis, opus streams and mp4 format for all other streams. > In this method the previous behavior of dashenc muxer could be > restored with little addition to the overall code. Also it's usage > will be simpler and easier to understand. > >>> > >>> This solution might be ok for just restoring the previous capability, > >>> but I think the ability for selecting the container format by the > >>> user is still more useful. For example, Opus can be muxed in both mp4 > >>> (although, with experimental flag) and webm, and it may make sense to > >>> some users to select mp4. (In my case, though, I wanted webm, hence > >>> the patch.) > >> In that case they could select "dash_segment_type" as "mp4", in which > >> case all streams including opus will be muxed in mp4 format. I don't > >> see a use-case where somebody wants opus in mp4 and would want other > >> streams in webm format. > > > > Suppose you want to create a DASH stream consisting of VP8, H264, Vorbis > > and Opus substreams to cover the best compatibility with clients (which > > are mostly web browsers, but not necessarilly all of them). AFAIK, you > > cannot put all these codecs neither in mp4 nor webm. An "auto" option > > would put Opus in webm container, leaving clients not supporting webm > > not able to play audio. With explicit container selection one could put > > Opus content in mp4 or both webm and mp4 (in which case the substream > > would be duplicated). > > > > An example of a client that is picky about container format is Safari on > > OS X High Sierra. I don't have one to test, but reportedly it supports > > Opus only in caf format, and I've read someone hacked it to play Opus in > > mp4 if disguised as AAC. I know lavf/dashenc doesn't support caf (yet) > > but it may support it in the future, so the container format selection > > would become even more relevant then. > > So, what do we decide about this patch? I understand your point. But all your examples are still theoretical(say caf is not supported in dashenc. And opus example doesn't mention any sample clients for that use-case). My view is that, if we can find one practical example where this advanced patch will be of immediate use(which cannot be done by auto option), then let us review and push this patch. If we can't find any immediate practical use-case, which could be tested and verified in popular clients then let's go for auto option. > > >>> Besides the parser, it doesn't add much code, and if I can improve > >>> the patch, please let me know how. If you absolutely don't want this > >>> functionality, that's ok, I'll keep this patch for my local builds > >>> and submit an "auto" option patch instead. > >> I am not absolutely against this patch. But I am just trying to find > >> if there is a use-case for such an advanced option. If there is a > >> practical requirement for such a use-case, then I agree that we should > >> review and push this patch for sure. > >> But if the requirement is just theoretical at this point, then I would > >> prefer the "auto" option patch. Maybe we could revisit this advanced > >> options patch when a real requirement comes up. > >>> > > --- > >doc/muxers.texi | 8 ++- > >libavformat/dashenc.c | 161 > > +++--- > >2 files changed, 140 insertions(+), 29 deletions(-) > > > > diff --git a/doc/muxers.texi b/doc/muxers.texi > > index 62f4091e31..4418b38c76 100644 > > --- a/doc/muxers.texi > > +++
Re: [FFmpeg-devel] [PATCH] avfilter: add astretch filter
Paul B Mahol (2018-11-14): > > libavfilter/Makefile | 1 + > > libavfilter/af_astretch.c | 330 ++ > > libavfilter/allfilters.c | 1 + > > 3 files changed, 332 insertions(+) > > create mode 100644 libavfilter/af_astretch.c > Will apply. You forgot the docs. Regards, -- Nicolas George signature.asc Description: Digital signature ___ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
Re: [FFmpeg-devel] [PATCH] hwcontext_vaapi: Add option to set driver name
On Wed, Nov 14, 2018 at 12:21:07AM +, Mark Thompson wrote:
> For example: -init_hw_device vaapi:/dev/dri/renderD128,driver=foo
>
> This may be more convenient that using the environment variable, and allows
> loading different drivers for different devices in the same process.
> ---
> libavutil/hwcontext_vaapi.c | 13 +
> 1 file changed, 13 insertions(+)
>
> diff --git a/libavutil/hwcontext_vaapi.c b/libavutil/hwcontext_vaapi.c
> index 8624369bb9..eb3a78d181 100644
> --- a/libavutil/hwcontext_vaapi.c
> +++ b/libavutil/hwcontext_vaapi.c
> @@ -1469,6 +1469,8 @@ static int vaapi_device_create(AVHWDeviceContext *ctx,
> const char *device,
> {
> VAAPIDevicePriv *priv;
> VADisplay display = NULL;
> +const AVDictionaryEntry *driver;
> +VAStatus vas;
>
> priv = av_mallocz(sizeof(*priv));
> if (!priv)
> @@ -1530,6 +1532,17 @@ static int vaapi_device_create(AVHWDeviceContext *ctx,
> const char *device,
> return AVERROR(EINVAL);
> }
>
> +driver = av_dict_get(opts, "driver", NULL, 0);
> +if (driver) {
> +vas = vaSetDriverName(display, driver->value);
this breaks build here on ubuntu:
CC libavutil/hwcontext_vaapi.o
libavutil/hwcontext_vaapi.c: In function ‘vaapi_device_create’:
libavutil/hwcontext_vaapi.c:1537:9: error: implicit declaration of function
‘vaSetDriverName’ [-Werror=implicit-function-declaration]
vas = vaSetDriverName(display, driver->value);
^
cc1: some warnings being treated as errors
make: *** [libavutil/hwcontext_vaapi.o] Error 1
make: Target `all' not remade because of errors.
[...]
--
Michael GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB
Avoid a single point of failure, be that a person or equipment.
signature.asc
Description: PGP signature
___
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
Re: [FFmpeg-devel] HEVC decoder for Raspberry Pi
Hi >Hi > >On Tue, Nov 13, 2018 at 03:52:18PM +, John Cox wrote: >> Hi >> >> I have been developing a hevc decoder for Raspberry Pi for some time >> now. As active development has now pretty much ceased and the code is >> believed stable it seems a good time to try presenting it to the group. >> >> You can find the current code on branch test/4.1.0/rpi_main in repo >> https://github.com/jc-kynesim/rpi-ffmpeg.git. It is based off tag n4.1 >> so if you diff it against n4.1 you should get a patch. >> >> This code has been in use by the Raspberry Pi version of Kodi for over >> two years now. >> >> If you think it would be a good idea to add this to the main ffmpeg >> distribution then I am willing to put reasonable effort into beating it >> into an appropriate shape. >> >> If not then it contains a reasonable number of ARM asm functions and >> other code that you might like to take/adapt for the current decoder. >> >> You will find the config scripts I have been using and a few notes in >> the pi-util directory if you wish to try building it for yourself. >> >> Just in case it isn't obvious: this will only run on a Pi. Slightly >> less obviously you need a Pi2 or better as the Pi0 & Pi1 don't have neon >> and are just too slow anyway. > >others may have other oppinions, but i think optimized code in FFmpeg >for Pi would be a good idea. >How to integrate this best though i do not know. And i cant know as >i have just quickly scrolled over the changes not really looked in detail Well if you want help with understanding what I've done feel free to email me and I'll do my best to explain. >But its certainly better to have hw optimizations in main git and >not have a seperate repository that needs to be maintained seperatly >for each platform ... and that the user has to find also ... and then >3rd party apps could have even more issues here if they wanted to use >optimized libs ... As I said I'm happy to put in reasonable amounts of work to make this happen. If we do want to go ahead then may I suggest that the most efficient way of proceeding would be that I take advice from one experienced person who understands the current hevc code (Michael?) by email until the work is mostly done and then return to the list for final polish? Regards JC ___ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
Re: [FFmpeg-devel] [PATCH v3 1/1] avcodec: implemented Media Foundation encoder wrapper
On Tue, Nov 13, 2018 at 6:52 PM Michael Niedermayer wrote: > On Tue, Nov 13, 2018 at 12:23:42PM +0100, Paweł Wegner wrote: > > On Tue, Nov 13, 2018 at 11:40 AM Michael Niedermayer > > > wrote: > > > > > On Tue, Nov 13, 2018 at 09:50:17AM +0100, Paweł Wegner wrote: > > > > On Tue, Nov 13, 2018 at 12:39 AM Michael Niedermayer > > > > > > > wrote: > > > > > > > > > On Mon, Nov 12, 2018 at 02:42:36PM +0100, Paweł Wegner wrote: > > > > > > Implemented the following encoders: > > > > > > * ff_h264_mf_encoder > > > > > > * ff_hevc_mf_encoder > > > > > > * ff_aac_mf_encoder > > > > > > * ff_ac3_mf_encoder > > > > > > * ff_mp3_mf_encoder > > > > > > > > > > > > The code is based on the PLEX's FFmpeg fork which can be found > here: > > > > > > https://files.plexapp.com/elan/ffmpeg/plex-ffmpeg-2017-12-17.txz > > > > > > > > > > > > Signed-off-by: Paweł Wegner > > > > > > --- > > > > > > configure | 11 + > > > > > > libavcodec/Makefile|1 + > > > > > > libavcodec/allcodecs.c |5 + > > > > > > libavcodec/mf_utils.c | 734 > > > > > > libavcodec/mf_utils.h | 207 +++ > > > > > > libavcodec/mfenc.c | 1242 > > > > > > > > > 6 files changed, 2200 insertions(+) > > > > > > create mode 100644 libavcodec/mf_utils.c > > > > > > create mode 100644 libavcodec/mf_utils.h > > > > > > create mode 100644 libavcodec/mfenc.c > > > > > > > > > > breaks build for mingw64 > > > > > > > > > > > > > CC libavcodec/mf_utils.o > > > > > In file included from > /usr/share/mingw-w64/include/wmcodecdsp.h:613:0, > > > > > from src/libavcodec/mf_utils.h:26, > > > > > from src/libavcodec/mf_utils.c:24: > > > > > /usr/share/mingw-w64/include/strmif.h:15380:2: warning: #warning > COM > > > > > interfaces layout in this header has not been verified. [-Wcpp] > > > > > #warning COM interfaces layout in this header has not been > verified. > > > > > ^ > > > > > /usr/share/mingw-w64/include/strmif.h:15381:2: warning: #warning > COM > > > > > interfaces with incorrect layout may not work at all. [-Wcpp] > > > > > #warning COM interfaces with incorrect layout may not work at all. > > > > > ^ > > > > > /usr/share/mingw-w64/include/strmif.h:15410:2: warning: #warning > COM > > > > > interfaces layout in this header has not been verified. [-Wcpp] > > > > > #warning COM interfaces layout in this header has not been > verified. > > > > > ^ > > > > > /usr/share/mingw-w64/include/strmif.h:15411:2: warning: #warning > COM > > > > > interfaces with incorrect layout may not work at all. [-Wcpp] > > > > > #warning COM interfaces with incorrect layout may not work at all. > > > > > ^ > > > > > > > > > I can silence these warnings by doing #undef __GNU_C__ before the MF > > > > inclusion; but would that really be better? > > > > This is where this comes from: > > > > > > > > https://github.com/Alexpux/mingw-w64/blob/master/mingw-w64-headers/include/mfplay.h#L59 > > > > > > > > > > > > > In file included from > /usr/share/mingw-w64/include/wmcodecdsp.h:613:0, > > > > > from src/libavcodec/mf_utils.h:26, > > > > > from src/libavcodec/mf_utils.c:24: > > > > > /usr/share/mingw-w64/include/strmif.h:15382:9: note: #pragma > message: > > > > > Interface IAMAsyncReaderTimestampScaling has unverified layout. > > > > > __MINGW_BROKEN_INTERFACE(INTERFACE) > > > > > ... > > > > > > > > > This one comes from mingw as well; and can be worked around the same > way. > > > > > > sorry i have cut the compiler output incorrectly > > > heres the full output, there are multiple real errors, not just > warnings: > > > > > > > > > > > > In file included from /usr/share/mingw-w64/include/wmcodecdsp.h:613:0, > > > from src/libavcodec/mf_utils.h:26, > > > from src/libavcodec/mf_utils.c:24: > > > /usr/share/mingw-w64/include/strmif.h:15380:2: warning: #warning COM > > > interfaces layout in this header has not been verified. [-Wcpp] > > > #warning COM interfaces layout in this header has not been verified. > > > ^ > > > /usr/share/mingw-w64/include/strmif.h:15381:2: warning: #warning COM > > > interfaces with incorrect layout may not work at all. [-Wcpp] > > > #warning COM interfaces with incorrect layout may not work at all. > > > ^ > > > /usr/share/mingw-w64/include/strmif.h:15410:2: warning: #warning COM > > > interfaces layout in this header has not been verified. [-Wcpp] > > > #warning COM interfaces layout in this header has not been verified. > > > ^ > > > /usr/share/mingw-w64/include/strmif.h:15411:2: warning: #warning COM > > > interfaces with incorrect layout may not work at all. [-Wcpp] > > > #warning COM interfaces with incorrect layout may not work at all. > > > ^ > > > In file included from /usr/share/mingw-w64/include/wmcodecdsp.h:613:0, > > > from src/libavcodec/mf_utils.h:26, > > > from
[FFmpeg-devel] [PATCH] avcodec/vc1: correct aspect ratio calculation
According to VC-1 spec:
* Display size defaults to max coded size when not explicitly set in
sequence header
* Aspect ratio in the sequence header refers to the Display size elements.
Therefore, the aspect ratio for the coded samples (SAR) needs to take into
account the scaling from coded size to display size, and the aspect ratio
of the display size elements.
Signed-off-by: Jerome Borsboom
---
VC-1 spec assumes that the output of the decoder, i.e. the pixel matrix with
dimensions coded_width x coded_height, is scaled to display size, i.e. a pixel
matrix of display_horiz_size x display_vert_size. This may introduce part of
the sample aspect ratio when the sizes of the two pixel matrices are not equal.
A further part of the sample aspect ratio is optionally specified in the
sequence
header as the aspect ratio of the display size pixels. This patch takes both
aspect ratios into account and aims to be correct even when the coded image
includes overscan regions.
libavcodec/vc1.c | 38 +-
libavcodec/vc1.h | 2 ++
2 files changed, 23 insertions(+), 17 deletions(-)
diff --git a/libavcodec/vc1.c b/libavcodec/vc1.c
index 3581d87b57..efc6edc4b0 100644
--- a/libavcodec/vc1.c
+++ b/libavcodec/vc1.c
@@ -442,30 +442,24 @@ static int decode_sequence_header_adv(VC1Context *v,
GetBitContext *gb)
}
v->s.max_b_frames = v->s.avctx->max_b_frames = 7;
if (get_bits1(gb)) { //Display Info - decoding is not affected by it
-int w, h, ar = 0;
+int ar = 0;
av_log(v->s.avctx, AV_LOG_DEBUG, "Display extended info:\n");
-w = get_bits(gb, 14) + 1;
-h = get_bits(gb, 14) + 1;
-av_log(v->s.avctx, AV_LOG_DEBUG, "Display dimensions: %ix%i\n", w, h);
+v->disp_horiz_size = get_bits(gb, 14) + 1;
+v->disp_vert_size = get_bits(gb, 14) + 1;
+av_log(v->s.avctx, AV_LOG_DEBUG, "Display dimensions: %ix%i\n",
+ v->disp_horiz_size, v->disp_vert_size);
if (get_bits1(gb))
ar = get_bits(gb, 4);
if (ar && ar < 14) {
-v->s.avctx->sample_aspect_ratio = ff_vc1_pixel_aspect[ar];
+v->aspect_ratio = ff_vc1_pixel_aspect[ar];
} else if (ar == 15) {
-w = get_bits(gb, 8) + 1;
-h = get_bits(gb, 8) + 1;
-v->s.avctx->sample_aspect_ratio = (AVRational){w, h};
+v->aspect_ratio = (AVRational){get_bits(gb, 8) + 1, get_bits(gb,
8) + 1};
} else {
-av_reduce(>s.avctx->sample_aspect_ratio.num,
- >s.avctx->sample_aspect_ratio.den,
- v->s.avctx->height * w,
- v->s.avctx->width * h,
- 1 << 30);
+v->aspect_ratio = (AVRational){1, 1};
}
-ff_set_sar(v->s.avctx, v->s.avctx->sample_aspect_ratio);
-av_log(v->s.avctx, AV_LOG_DEBUG, "Aspect: %i:%i\n",
- v->s.avctx->sample_aspect_ratio.num,
- v->s.avctx->sample_aspect_ratio.den);
+av_log(v->s.avctx, AV_LOG_DEBUG, "Aspect ratio: %i:%i\n",
+ v->aspect_ratio.num,
+ v->aspect_ratio.den);
if (get_bits1(gb)) { //framerate stuff
if (get_bits1(gb)) {
@@ -490,6 +484,10 @@ static int decode_sequence_header_adv(VC1Context *v,
GetBitContext *gb)
v->transfer_char = get_bits(gb, 8);
v->matrix_coef = get_bits(gb, 8);
}
+} else {
+v->disp_horiz_size = v->max_coded_width;
+v->disp_vert_size = v->max_coded_height;
+v->aspect_ratio = (AVRational){1, 1};
}
v->hrd_param_flag = get_bits1(gb);
@@ -544,6 +542,12 @@ int ff_vc1_decode_entry_point(AVCodecContext *avctx,
VC1Context *v, GetBitContex
av_log(avctx, AV_LOG_ERROR, "Failed to set dimensions %d %d\n", w, h);
return ret;
}
+av_reduce(>sample_aspect_ratio.num,
+ >sample_aspect_ratio.den,
+ v->disp_horiz_size * v->aspect_ratio.num * h,
+ v->disp_vert_size * v->aspect_ratio.den * w,
+ 1 << 30);
+ff_set_sar(avctx, avctx->sample_aspect_ratio);
if (v->extended_mv)
v->extended_dmv = get_bits1(gb);
diff --git a/libavcodec/vc1.h b/libavcodec/vc1.h
index 69f6ca9e4d..7674b0f9a1 100644
--- a/libavcodec/vc1.h
+++ b/libavcodec/vc1.h
@@ -209,6 +209,8 @@ typedef struct VC1Context{
int hrd_param_flag; ///< Presence of Hypothetical Reference
///< Decoder parameters
int psf; ///< Progressive Segmented Frame
+int disp_horiz_size, disp_vert_size;
+AVRational aspect_ratio;
//@}
/** Sequence header data for all Profiles
--
2.13.6
___
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
Re: [FFmpeg-devel] [PATCH] lavf/dashenc: set FLAC manifest codec string to "flac"
On 11/14/18, Jan Ekström wrote: > Internally in ISOBMFF the FLAC-in-ISOBMFF draft uses the "fLaC" > as the identifier for FLACSampleEntry, but what the browsers > seem to eat for MPEG-DASH manifests is plain "flac". > > Verified by auri_ on IRC to play with the major browsers. > --- > libavformat/dashenc.c | 1 + > 1 file changed, 1 insertion(+) > lgtm ___ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
[FFmpeg-devel] [PATCH] avcodec/diracdec: Check component quant
Fixes: Timeout
Fixes:
10708/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_DIRAC_fuzzer-5730140957442048
Found-by: continuous fuzzing process
https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg
Signed-off-by: Michael Niedermayer
---
libavcodec/diracdec.c | 5 +
1 file changed, 5 insertions(+)
diff --git a/libavcodec/diracdec.c b/libavcodec/diracdec.c
index af561d1426..fd23139062 100644
--- a/libavcodec/diracdec.c
+++ b/libavcodec/diracdec.c
@@ -676,6 +676,11 @@ static int decode_component(DiracContext *s, int comp)
b->length = get_interleaved_ue_golomb(>gb);
if (b->length) {
b->quant = get_interleaved_ue_golomb(>gb);
+if (b->quant > (DIRAC_MAX_QUANT_INDEX - 1)) {
+av_log(s->avctx, AV_LOG_ERROR, "Unsupported quant %d\n",
b->quant);
+b->quant = 0;
+return AVERROR_INVALIDDATA;
+}
align_get_bits(>gb);
b->coeff_data = s->gb.buffer + get_bits_count(>gb)/8;
b->length = FFMIN(b->length, FFMAX(get_bits_left(>gb)/8,
0));
--
2.19.1
___
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
Re: [FFmpeg-devel] [PATCH] avfilter: add astretch filter
On 11/10/18, Paul B Mahol wrote: > Signed-off-by: Paul B Mahol > --- > libavfilter/Makefile | 1 + > libavfilter/af_astretch.c | 330 ++ > libavfilter/allfilters.c | 1 + > 3 files changed, 332 insertions(+) > create mode 100644 libavfilter/af_astretch.c Will apply. ___ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
Re: [FFmpeg-devel] [PATCH] lavf/dashenc: Add support for per-stream container type selection.
On 11/12/18 3:55 PM, Andrey Semashev wrote:
On 11/12/18 3:12 PM, Jeyapal, Karthick wrote:
On 11/12/18 5:20 PM, Andrey Semashev wrote:
On 11/12/18 8:20 AM, Jeyapal, Karthick wrote:
On 11/8/18 10:27 PM, Andrey Semashev wrote:
This commit restores the ability to create DASH streams with codecs
that require different containers that was lost after commit
2efdbf7367989cf9d296c25fa3d2aff8d6e25fdd. It extends the
dash_segment_type
option syntax to allow to specify segment container types for
individual
streams, in addition to the default container type that is applied to
all streams. The extended syntax is backward compatible with the
previous
syntax.
Thanks for sending the patch. I understand your requirement completely.
But I feel that this option for mapping streams with container
format is little confusing. Also, the relevant code is relatively
big, and thus difficult to maintain in future.
I have a middle ground suggestion. If your goal is to achieve the
earlier behavior broken commits, then I propose the following.
Option "dash_segment_type" could take one more option "auto"
(instead of mp4 or webm).
When "auto" is chosen, the muxer could choose webm format for VP8,
VP9, vorbis, opus streams and mp4 format for all other streams.
In this method the previous behavior of dashenc muxer could be
restored with little addition to the overall code. Also it's usage
will be simpler and easier to understand.
This solution might be ok for just restoring the previous capability,
but I think the ability for selecting the container format by the
user is still more useful. For example, Opus can be muxed in both mp4
(although, with experimental flag) and webm, and it may make sense to
some users to select mp4. (In my case, though, I wanted webm, hence
the patch.)
In that case they could select "dash_segment_type" as "mp4", in which
case all streams including opus will be muxed in mp4 format. I don't
see a use-case where somebody wants opus in mp4 and would want other
streams in webm format.
Suppose you want to create a DASH stream consisting of VP8, H264, Vorbis
and Opus substreams to cover the best compatibility with clients (which
are mostly web browsers, but not necessarilly all of them). AFAIK, you
cannot put all these codecs neither in mp4 nor webm. An "auto" option
would put Opus in webm container, leaving clients not supporting webm
not able to play audio. With explicit container selection one could put
Opus content in mp4 or both webm and mp4 (in which case the substream
would be duplicated).
An example of a client that is picky about container format is Safari on
OS X High Sierra. I don't have one to test, but reportedly it supports
Opus only in caf format, and I've read someone hacked it to play Opus in
mp4 if disguised as AAC. I know lavf/dashenc doesn't support caf (yet)
but it may support it in the future, so the container format selection
would become even more relevant then.
So, what do we decide about this patch?
Besides the parser, it doesn't add much code, and if I can improve
the patch, please let me know how. If you absolutely don't want this
functionality, that's ok, I'll keep this patch for my local builds
and submit an "auto" option patch instead.
I am not absolutely against this patch. But I am just trying to find
if there is a use-case for such an advanced option. If there is a
practical requirement for such a use-case, then I agree that we should
review and push this patch for sure.
But if the requirement is just theoretical at this point, then I would
prefer the "auto" option patch. Maybe we could revisit this advanced
options patch when a real requirement comes up.
---
doc/muxers.texi | 8 ++-
libavformat/dashenc.c | 161
+++---
2 files changed, 140 insertions(+), 29 deletions(-)
diff --git a/doc/muxers.texi b/doc/muxers.texi
index 62f4091e31..4418b38c76 100644
--- a/doc/muxers.texi
+++ b/doc/muxers.texi
@@ -289,8 +289,12 @@ Set container format (mp4/webm) options using
a @code{:} separated list of
key=value parameters. Values containing @code{:} special
characters must be
escaped.
-@item dash_segment_type @var{dash_segment_type}
-Possible values:
+@item -dash_segment_type @var{dash_segment_type}
+Sets the container type for dash segment files. Syntax is "
:a,b,c :d,e" where is
+the container type and a, b, c, d and e are the indices of the
mapped streams. When no indices are specified,
+the container type is set for all streams.
+
+Possible container type values:
@item mp4
If this flag is set, the dash segment files will be in in
ISOBMFF format. This is the default format.
diff --git a/libavformat/dashenc.c b/libavformat/dashenc.c
index f8b3d106d5..626dc76413 100644
--- a/libavformat/dashenc.c
+++ b/libavformat/dashenc.c
@@ -84,6 +84,8 @@ typedef struct OutputStream {
int64_t first_pts, start_pts, max_pts;
int64_t last_dts, last_pts;
