We have no valid sample that triggers this so we do not know if this would decode
correctly, but -inf is not the correct k value
Fixes: Assertion n>=0 && n<=32 failed at libavcodec/get_bits.h:423
Fixes: -inf is outside the range of representable values of type 'int'
Fixes: 70709/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_OSQ_fuzzer-6223623839350784
Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg
While shifting "value" to left, we would iterate through all bits
of an unsigned long long, while we only expect to count through
"size * CHAR_BIT" bits; instead shift bits to the right and just
count the trailing zeros.
This fixes fate with MSVC.
Signed-off-by: Martin Storsjö <martin@martin.st>
Since c0666d8b, rgb24toyv12 is broken for width non-aligned to 16.
Add a simple wrapper to handle the non-aligned part.
Co-authored-by: johzzy <hellojinqiang@gmail.com>
Signed-off-by: Zhao Zhili <zhilizhao@tencent.com>
Only those that can be decoded with our implementation, so excluding
* C and D - independent layers
* G, H, I - more than 2 layers
Frame hashes verified against the reference implementation from
https://hevc.hhi.fraunhofer.de/svn/svn_3DVCSoftware/
This extends the syntax for specifying input streams in -map and complex
filtergraph labels, to allow selecting a view by view ID, index, or
position. The corresponding decoder is then set up to decode the
appropriate view and send frames for that view to the correct
filtergraph input(s).
The HEVC decoder will start setting stereoscopic view position (left or
right) based on 3D Reference Displays Info SEI message in future
commits. This information should be merged with container-derived
stereo3D side data.
Only do it in reget_buffer().
The purpose of this clearing this flag is to prevent it for
unintentionally persisting across multiple invocations of this function
on one frame, however that is only a problem if the frame is not
unreffed between uses, which is only the case with reget_buffer().
In other cases the caller may legitimately want to set the discard flag
and should have the option of doing so.
The per-frame reference picture set contains two more lists -
INTER_LAYER[01]. Assuming at most two layers, INTER_LAYER1 is always
empty, but is added anyway for completeness.
When inter-layer prediction is enabled, INTER_LAYER0 for the
second-layer frame will contain the base-layer frame from the same
access unit, if it exists.
The new lists are then used in per-slice reference picture set
construction as per F.8.3.4 "Decoding process for reference picture
lists construction".
I encountered this problem with NTP timestamps that are extremely old,
like from January, 1990.
Although RFC3550 suggests that the timestamps in the RTCP packets use
the actual wallclock, some implementations use other clocks, such as
the CLOCK_MONOTONIC on linux.
I'm my case, I'm dealing with packets from mediasoup.
Without this patch, start_time_realtime shows up in the distance future
instead of around Jan 1900.
Signed-off-by: Anton Khirnov <anton@khirnov.net>
If not, report it as a bug. avfilter_graph_create_filter() will return ENOMEM if the
passed filter argument is NULL, which is misleading.
Signed-off-by: James Almer <jamrial@gmail.com>
While running this command
./ffmpeg_g -loglevel debug -hwaccel vulkan -init_hw_device vulkan=vk:0,debug=1 -hwaccel_output_format vulkan -i input.y4m -vf 'format=nv12,hwupload' -c:v h264_vulkan -quality 2 output.mp4 -y
It hit this validation error:
Validation Error: [ VUID-vkCmdEncodeVideoKHR-None-08318 ] Object 0: handle =
0x8f000000008f, type = VK_OBJECT_TYPE_VIDEO_SESSION_KHR; Object 1: handle =
0xfd00000000fd, type = VK_OBJECT_TYPE_VIDEO_SESSION_PARAMETERS_KHR;
| MessageID = 0x5dc3dd39
| vkCmdEncodeVideoKHR(): The currently configured encode quality level (2) for
VkVideoSessionKHR 0x8f000000008f[] does not match the encode quality level (0)
VkVideoSessionParametersKHR 0xfd00000000fd[] was created with. The Vulkan spec
states: The bound video session parameters object must have been created with
the currently set video encode quality level for the bound video session at the
time the command is executed on the
device (https://www.khronos.org/registry/vulkan/specs/1.3-extensions/html/vkspec.html#VUID-vkCmdEncodeVideoKHR-None-08318)
This patch adds a new function helper for creating session parameters, which
also sets the quality level and it's called by the H.264 and H.265 Vulkan
encoders.
Push descriptors are in theory slightly faster, but come with
limitations for which we have to check.
Either way, they're not difficult to implement, so even though
no one should be using peasant-tier descriptors, do it anyway.
This permits:
- The use of Vulkan filtering on many more devices
- Better debugging due to lack of descriptor buffer support in layers
Much of the changes here are due to a requirement that updates to
descriptors must happen between the command buffer being waited on,
and the pipeline not being bound.
We routinely did it the other way around, by updating only after
we bind the pipeline.
More recent kernel versions allow for users to extract a sync_file
handle from a DMA-BUF, which can then be imported into Vulkan as a
binary semaphore.
This finally allows for synchronization between Vulkan and DMA-BUF
images, such as those from screen capture software, or VAAPI,
avoiding any corruption artifacts.
This is done fully asynchronously, where we use the kernel's
given binary semaphores as a dependency to increment the image's
usual VkSemaphores we allocate. The old imported binary semaphores
are cleaned up after execution as usual.
In the future, hwcontext_drm should receive support for explicitly
synchronized images as well, which would make the synchronization
more robust and portable.
