This commit adds a common code for use in Vulkan filters. It attempts
to ease the burden of writing Vulkan image filtering to a minimum,
which is pretty much a requirement considering how verbose the API is.
It supports both compute and graphic pipelines and manages to abstract
the API to such a
On 5/21/2018 11:46 PM, Rostislav Pehlivanov wrote:
> This commit adds a common code for use in Vulkan filters. It attempts
> to ease the burden of writing Vulkan image filtering to a minimum,
> which is pretty much a requirement considering how verbose the API is.
>
> It supports both compute and
Can convert to RGB using very fast fixed-function conversions.
Signed-off-by: Rostislav Pehlivanov
---
configure | 1 +
libavfilter/Makefile | 1 +
libavfilter/allfilters.c | 1 +
libavfilter/vf_scale_vulkan.c | 395
Could be done in-plane with the main image but framesync segfaults.
Signed-off-by: Rostislav Pehlivanov
---
configure | 1 +
libavfilter/Makefile| 1 +
libavfilter/allfilters.c| 1 +
libavfilter/vf_overlay_vulkan.c | 461
This commit adds a common code for use in Vulkan filters. It attempts
to ease the burden of writing Vulkan image filtering to a minimum,
which is pretty much a requirement considering how verbose the API is.
It supports both compute and graphic pipelines and manages to abstract
the API to such a
It tries to do something similar to it with YUV images, but RGB images
are done properly.
Signed-off-by: Rostislav Pehlivanov
---
configure | 1 +
libavfilter/Makefile| 1 +
libavfilter/allfilters.c
Signed-off-by: Rostislav Pehlivanov
---
configure | 1 +
libavfilter/Makefile| 1 +
libavfilter/allfilters.c| 1 +
libavfilter/vf_avgblur_vulkan.c | 343
4 files changed, 346 insertions(+)
Used to fix unmapping when no direct interop exists between APIs.
Signed-off-by: Rostislav Pehlivanov
---
libavutil/hwcontext.c | 7 +++
libavutil/hwcontext_internal.h | 5 +
2 files changed, 12 insertions(+)
diff --git a/libavutil/hwcontext.c
Signed-off-by: Rostislav Pehlivanov
---
libavutil/hwcontext_opencl.c | 5 +
1 file changed, 1 insertion(+), 4 deletions(-)
diff --git a/libavutil/hwcontext_opencl.c b/libavutil/hwcontext_opencl.c
index 43b5c5ae0c..1d18da37bf 100644
--- a/libavutil/hwcontext_opencl.c
+++
This commit adds a Vulkan hwcontext, currently capable of mapping DRM and
VAAPI frames but additional functionality can be added later to support
importing of D3D11 surfaces as well as exporting to various other APIs.
This context requires the newest stable version of the Vulkan API,
and once the
This is pretty much finished.
Rostislav Pehlivanov (8):
hwcontext_internal: add ff_hwframe_map_replace
hwcontext_opencl: use ff_hwframe_map_replace()
lavu: add a Vulkan hwcontext
lavfi: add common Vulkan filtering code
lavfi: add a Vulkan avgblur filter
lavfi: add a Vulkan chromatic
On Tue, 22 May 2018 01:18:30 +0100, Mark Thompson wrote:
> On 21/05/18 07:50, Ruiling Song wrote:
> > This filter does HDR(HDR10/HLG) to SDR conversion with tone-mapping.
> >
> > An example command to use this filter with vaapi codecs:
> > FFMPEG -init_hw_device
> -Original Message-
> From: myp...@gmail.com [mailto:myp...@gmail.com]
> Sent: Monday, May 21, 2018 3:23 PM
> To: FFmpeg development discussions and patches
> Cc: s...@jkqxz.net; ffm...@haasn.xyz; Song, Ruiling
> Subject: Re:
On 21/05/18 07:50, Ruiling Song wrote:
> This filter does HDR(HDR10/HLG) to SDR conversion with tone-mapping.
>
> An example command to use this filter with vaapi codecs:
> FFMPEG -init_hw_device vaapi=va:/dev/dri/renderD128 -init_hw_device \
> opencl=ocl@va -hwaccel vaapi -hwaccel_device va
For lack of a better option here's a dropbox link. The
ftp://upload.ffmpeg.org server doesn't appear to respond. This is a tiny
clip unmodified from NVidia Shadowplay.
(5.4MB)
https://www.dropbox.com/s/6izrx87gfsgjodq/mov-trak-name-shadowplay.mp4?dl=0
-- Original Message --
From:
Hi!
Attached patch allows pal8 encoding into j2k, tested with libopenjpeg,
kakadu and jasper.
Please comment, Carl Eugen
From ffc9968279ecbf8b0b6659e9a07539edb9787e44 Mon Sep 17 00:00:00 2001
From: Carl Eugen Hoyos
Date: Tue, 22 May 2018 02:21:04 +0200
Subject: [PATCH]
2018-05-21 16:18 GMT+02:00, Nik Johnson :
> Some muxers produce mp4s with a udta (user data) atom nested within a
> trak atom. The nested udta atoms typically contain stream information
> such as the title. ffmpeg should treat nested udta atoms as applicable
> to the stream
On 5/21/2018 3:16 PM, Gagandeep Singh wrote:
> ---
> libavcodec/cfhd.c | 177
> +-
> libavcodec/cfhd.h | 9 +++
> 2 files changed, 158 insertions(+), 28 deletions(-)
>
> diff --git a/libavcodec/cfhd.c b/libavcodec/cfhd.c
> index
2018-05-21 20:16 GMT+02:00, Gagandeep Singh :
> ---
> libavcodec/cfhd.c | 177
> +-
> libavcodec/cfhd.h | 9 +++
> 2 files changed, 158 insertions(+), 28 deletions(-)
Please mention ticket #5522 in the commit
On 5/21/2018 4:16 PM, James Zern wrote:
> On Sat, May 19, 2018 at 11:16 AM James Almer wrote:
>
>> The libvpx doxy says that a value of 0 for the g_threads field is
>> equivalent to a value of 1, whereas for avctx->thread_count it means
>> the maximum amount of threads
On 5/21/2018 11:15 AM, James Almer wrote:
> On 5/21/2018 8:27 AM, Thomas Volkert wrote:
>> Hi,
>>
>> On 14.05.2018 19:04, Dave Gregory wrote:
>>> Hi all,
>>>
>>> [..comparison with alternative TLS implementations..]
>>
>> Ping.
>>
>> I still suggest to support mbedTLS in FFmpeg because it is
This can happen if frames_init fails.
---
libavutil/hwcontext_opencl.c | 10 ++
1 file changed, 6 insertions(+), 4 deletions(-)
diff --git a/libavutil/hwcontext_opencl.c b/libavutil/hwcontext_opencl.c
index c29a521f07..0d255e54e7 100644
--- a/libavutil/hwcontext_opencl.c
+++
This ensures that the format and dimensions are actually valid -
previously this wasn't checked, so we would only fail when the user
actually attempted to allocate an image.
---
libavutil/hwcontext_opencl.c | 12
1 file changed, 12 insertions(+)
diff --git
Creates a frames context of every available type and attempts to
upload/download frame data to each, checking that the output is the same
as the input in each case.
---
This test passes with VDPAU and OpenCL, but may fail for others.
For VAAPI the main issue is that there isn't really a
The code previously assumed that src->hw_frames_ctx was necessarily set
if dst->buf wasn't. Using non-refcounted hardware frames is not allowed
here, but we can return an error in that case.
---
libavutil/hwcontext.c | 8 ++--
1 file changed, 6 insertions(+), 2 deletions(-)
diff --git
On Mon, May 21, 2018 at 02:36:28PM -0700, Wang Cao wrote:
> Sorry the ticket doesn't contain alaw_2.aif as in
> https://trac.ffmpeg.org/ticket/1660
the issue is reproduceable with alaw.aif from the ticket
[...]
--
Michael GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB
Many
On 16/05/18 08:19, Haihao Xiang wrote:
> In lavc/hevec_vaapi, colour properties in AVCodecContext are needed to
> write the sequence header
>
> Tested by the command below:
> ffmpeg -hwaccel vaapi -hwaccel_device /dev/dri/renderD128
> -hwaccel_output_format vaapi -i input-with-hdr.mkv -c:v
On 16/05/18 02:14, Xiang, Haihao wrote:
> On Tue, 2018-05-15 at 23:15 +0100, Mark Thompson wrote:
>> This uses any devices it can find on the host system - on a system with no
>> hardware device support or in builds with no support included it will do
>> nothing and pass.
>> ---
>>
>> It now comes
Could you elaborate a bit on what you mean?
This patch mostly follows the pattern of the existing 'gmpd' track, however
it seems without adding logic to mov.c::mov_parse_stsd_data() the track is
not correctly identified. I just tested -c copy on a file with gpmd track
and it is lost during the
Sorry the ticket doesn't contain alaw_2.aif as in
https://trac.ffmpeg.org/ticket/1660
On Fri, May 18, 2018 at 8:13 PM, Michael Niedermayer wrote:
> On Fri, May 18, 2018 at 04:54:25PM -0700, Wang Cao wrote:
> > - Make ffmpeg to output stats for each video streams and
Fixes truncation
Fixes Assertion n <= 31 && value < (1U << n) failed at libavcodec/put_bits.h:169
Fixes: ffmpeg_crash_2.avi
Found-by: Thuan Pham , Marcel Böhme, Andrew Santosa
and Alexandru RazvanCaciulescu with AFLSmart
Signed-off-by: Michael Niedermayer
On Mon, May 21, 2018 at 3:04 PM, James Darnley wrote:
> On 2018-05-20 20:53, Stephan Holljes wrote:
>> +#include
>> +#include
>> +#include
>
> That's not portable. Lua headers are not in a subdirectory.
Yes, artifact from early testing, changed and tested, pkg-config
>
> +static inline void interlaced_vertical_filter(int16_t *output, int16_t
> *low, int16_t *high,
> + int width, int linesize, int plane)
> +{
> +int i;
> +int16_t even, odd;
> +for (i = 0; i < width; i++) {
> + even = (*low - *high)/2;
> + odd =
Not really "deinterlacing", it would be "add interlaced file support" or
similar.
On Mon, 21 May 2018 at 20:17 Gagandeep Singh
wrote:
> +static inline void peak_table(int16_t *band, Peak *peak, int length)
> +{
> +int i;
> +for (i = 0; i < length; i++)
> +
On Sun, May 20, 2018 at 12:55 PM, Michael Niedermayer <
mich...@niedermayer.cc> wrote:
> On Sun, May 20, 2018 at 11:37:22AM -0700, Aman Gupta wrote:
> > On Sat, May 19, 2018 at 2:56 PM James Almer wrote:
> >
> > > On 5/19/2018 6:31 PM, Michael Niedermayer wrote:
> > > > On
On 21 May 2018 at 19:52, Erik Ackermann wrote:
> Spec: https://developers.google.com/streetview/publish/camm-spec
>
> The Lenovo Mirage camera launched recently and outputs this metadata track
> in its videos.
>
> ___
>
On Sun, May 20, 2018 at 11:09 PM, Łukasz Krzciuk wrote:
> Any updates on this issue?
>
I believe that this patch was already merged:
https://github.com/FFmpeg/FFmpeg/commit/48330500efd636b1540002b600257d8802badc69
>
> Regards,
>
> *Łukasz Krzciuk*
> Developer
>
> Vewd
> ul.
On Sat, May 19, 2018 at 11:16 AM James Almer wrote:
> The libvpx doxy says that a value of 0 for the g_threads field is
> equivalent to a value of 1, whereas for avctx->thread_count it means
> the maximum amount of threads possible for the host system.
> Use av_cpu_count() to
On Mon, May 21, 2018 at 2:58 AM, Ruiling Song
wrote:
> If the transfer was SMPTE2084, use the peak of 1 even if not tagged.
> Otherwise, we would assume it is HLG with a peak of 1200.
> Based on suggestion by Niklas Haas.
>
> Signed-off-by: Ruiling Song
Some muxers write the stream title in a udta atom with the tag 'name'.
Recognize 'name' tags as the stream title instead of an unknown tag.
Signed-off-by: Nik Johnson
---
[v2: Add raw = 1]
libavformat/mov.c | 1 +
1 file changed, 1 insertion(+)
diff --git
Some muxers produce mp4s with a udta (user data) atom nested within a
trak atom. The nested udta atoms typically contain stream information
such as the title. ffmpeg should treat nested udta atoms as applicable
to the stream instead of globally.
Signed-off-by: Nik Johnson
Spec: https://developers.google.com/streetview/publish/camm-spec
The Lenovo Mirage camera launched recently and outputs this metadata track
in its videos.
From 937a49d00139edfc8cc5eda9234c2b8afbaa4773 Mon Sep 17 00:00:00 2001
From: Erik Ackermann
Date: Mon, 21 May 2018 11:42:04
---
libavcodec/cfhd.c | 177 +-
libavcodec/cfhd.h | 9 +++
2 files changed, 158 insertions(+), 28 deletions(-)
diff --git a/libavcodec/cfhd.c b/libavcodec/cfhd.c
index 7ceb803595..d70c13592c 100644
--- a/libavcodec/cfhd.c
+++
On Mon, May 21, 2018 at 10:29 AM Alexander Ivash wrote:
> I'm not injecting metadata, but reading it! Here:
>
> https://github.com/ElderOrb/qctools/blob/master/Source/Core/VideoStats.cpp#L287
> This function is being called for every frame. The issue is that the
> first frame
I'm not injecting metadata, but reading it! Here:
https://github.com/ElderOrb/qctools/blob/master/Source/Core/VideoStats.cpp#L287
This function is being called for every frame. The issue is that the
first frame returns null.
2018-05-21 19:35 GMT+03:00 Paul B Mahol :
> On
On 5/21/18, Alexander Ivash wrote:
> What could be the reason for such a behavior besides using deprecated
> 'avcodec_decode_video2' ? I've also tried to experiment a bit with new
> API (avcodec_send_packet/avcodec_receive_frame) but this time got no
> metadata even for second
On Mon, May 14, 2018 at 4:49 PM, Jacob Trimble wrote:
> On Tue, May 8, 2018 at 3:47 PM, Michael Niedermayer
> wrote:
>> On Mon, May 07, 2018 at 04:59:33PM -0700, Jacob Trimble wrote:
>>> On Mon, May 7, 2018 at 3:18 PM, Michael Niedermayer
>>>
What could be the reason for such a behavior besides using deprecated
'avcodec_decode_video2' ? I've also tried to experiment a bit with new
API (avcodec_send_packet/avcodec_receive_frame) but this time got no
metadata even for second frame (the code is here:
Some muxers write the stream title in a udta atom with the tag 'name'.
Recognize 'name' tags as the stream title instead of an unknown tag.
Signed-off-by: Nik Johnson
---
libavformat/mov.c | 1 +
1 file changed, 1 insertion(+)
diff --git a/libavformat/mov.c
Some muxers produce mp4s with a udta (user data) atom nested within a
trak atom. The nested udta atoms typically contain stream information
such as the title. ffmpeg should treat nested udta atoms as applicable
to the stream instead of globally.
Signed-off-by: Nik Johnson
On 5/21/2018 8:27 AM, Thomas Volkert wrote:
> Hi,
>
> On 14.05.2018 19:04, Dave Gregory wrote:
>> Hi all,
>>
>> [..comparison with alternative TLS implementations..]
>
> Ping.
>
> I still suggest to support mbedTLS in FFmpeg because it is especially
> interesting for embedded devices.
>
> Best
On 2018-05-20 20:53, Stephan Holljes wrote:
> +#include
> +#include
> +#include
That's not portable. Lua headers are not in a subdirectory.
> +int configs_read(struct HTTPDConfig **configs, const char *filename)
> +{
> +int ret = 0;
> +int nb_configs = 0;
> +int nb_streams = 0;
>
Hi,
On 14.05.2018 19:04, Dave Gregory wrote:
> Hi all,
>
> [..comparison with alternative TLS implementations..]
Ping.
I still suggest to support mbedTLS in FFmpeg because it is especially
interesting for embedded devices.
Best regards,
Thomas.
___
RGB32(AV_PIX_FMT_BGRA on intel platforms) format may be used as overlay with
alpha blending.
So add AV_PIX_FMT_BGRA format support.
Rename RGB32 to be BGRA to make it clearer as Mark Thompson's suggestion.
Signed-off-by: Zhong Li
---
libavfilter/qsvvpp.c | 2 +-
2018-05-21 14:50 GMT+08:00 Ruiling Song :
> This filter does HDR(HDR10/HLG) to SDR conversion with tone-mapping.
>
> An example command to use this filter with vaapi codecs:
> FFMPEG -init_hw_device vaapi=va:/dev/dri/renderD128 -init_hw_device \
> opencl=ocl@va -hwaccel
If the transfer was SMPTE2084, use the peak of 1 even if not tagged.
Otherwise, we would assume it is HLG with a peak of 1200.
Based on suggestion by Niklas Haas.
Signed-off-by: Ruiling Song
---
libavfilter/vf_tonemap.c | 5 ++---
1 file changed, 2 insertions(+), 3
This filter does HDR(HDR10/HLG) to SDR conversion with tone-mapping.
An example command to use this filter with vaapi codecs:
FFMPEG -init_hw_device vaapi=va:/dev/dri/renderD128 -init_hw_device \
opencl=ocl@va -hwaccel vaapi -hwaccel_device va -hwaccel_output_format \
vaapi -i INPUT
Not convenient if using numerals to set MFE mode. It is ambiguous
and misleading (e.g: user may misunderstand setting mfmode to 1 is to
enable MFE but actually it is to disable MFE, and set it to be 5 or above is
meaningless).
V2: remove the manual option since it is not supported now.
Any updates on this issue?
Regards,
*Łukasz Krzciuk*
Developer
Vewd
ul. Grabarska 2, Pegaz 2A, 50-079 Wrocław, Polska
On Tue, May 15, 2018 at 12:51 PM, Łukasz Krzciuk wrote:
> I have prepared a testcase. PMT data comes from org.hbbtv_HTML50420 - I
> have squashed them and
59 matches
Mail list logo