Our encoder implementation actually supports a small subset of the
formats supported by the decoder. Those are the formats for which we
have a copy path in gst_omx_video_enc_fill_buffer() and which are not
filtered out in filter_supported_formats().
Coverity was complaining with:
Null pointer dereferences (REVERSE_INULL) Null-checking "frame"
suggests that it may be null, but it has already been
dereferenced on all paths leading to the check.
The frame == NULL has been removed as 'frame' is actively used
in the code above without any change of dereferencing and setting
its value to NULL before the test.
CID: 1461287
Does not change anything for now but will be needed when we'll support
interlace-mode=alternate as the field rate will be twice the frame rate.
Made the code safe from division by 0 while I was on it.
The generic target is meant to only test building gst-omx. It doesn't
provide any configuration file and so is not supposed to register any
element.
I'm not aware of any user building gst-omx with this target and
providing their own conf file to actually register elements. But best to
not break this use case anyway so let's just downgrade the log message.
Fix GST_ERROR in the 'check fedora' CI job.
This patch adds look-ahead property to encoder
The value indicates look ahead size in frames,
the number of frames processed ahead of second pass encoding.
Dual pass encoding is disabled if look-ahead
value is less than 2.
Custom API that upstream elements can use to notify encoders about
marking longterm ref. pictures or using longterm ref. pictures in
encoding process.
This patch adds below properties:
long-term-ref: Enable/Disable dynamically marking long-term
reference pictures in encoding process
long-term-freq: Periodicity of long-term reference picture
marking in encoding process.
If a picture is marked as long-term reference picture then it remains
in the DPB list for ever unless it overrides with new long-term pitcure with
same index. Encoder can use this long-term picture as refence for
encoding.
This feature is mostly useful to avoid visual artifacts propagation in streaming use cases
when packet loss happens. Instead of requesting for IDR, client can request for use long-term
reference picture for encoding.
We are operating in stream-format=byte-stream so the codec data buffer
is meant to be part of the buffer flow.
The base class will push it when a key frame is requested (as we stored
it with gst_video_encoder_set_headers()) but we still have to push it
right away as part of the normal buffer flow.
Also set the HEADER flag on this buffer.
We now negotiate subframe mode through the caps. To enabled subframe
mode, the caps need to specify alignment=nal:
... ! omxh264enc ! video/x-h264,alignment=nal ! ...
... ! omxh265enc ! video/x-h265,alignment=nal ! ...
By passing the expected video buffer layout, the upstream producer
may be able to produce buffers fitting those requierements allowing
gst-omx to use dynamic buffer mode rather than having to copy each input
buffer.
This is particularly useful with v4l2src as it can request the capture
driver to produce buffers with the required paddings.
Tell buffer consumer about our paddings.
v4l2src can now uses these paddings information when trying to import
buffers to configure the v4l2 driver accordingly.
Fixing a regression introduced in my previous patch
(7c40a91c31).
The ALLOCATION query needs to be handled by GstVideoEncoder (to call
propose_allocation()) so chain up the query handling rather than early
returning.
Ensure that the encoder releases all its input buffers when requested by
upstream. Encoder input buffers may be shared with downstreaming (when
using dmabuf), upstream may then request the encoder to
drain when reconfiguring before destroying its buffers.
Also drain on ALLOCATION query as we already do in kmssink as that
notify of a format change.
Fix "decoder ! encoder" pipeline when decoding a file with different
resolutions on Zynq.
When importing dmabuf from downstream, we want the allocator to be in
OTHER_POOL mode despite output_mode being DMABUF.
So check first if other_pool is set before checking for pool's
output_mode.
Those debug infos have proved to be very helpful when debugging
timestamp issues. They are often linked to gst-omx picking the wrong
frame when trying to map from OMX.
One big restriction of the OMX buffer pool has always been
that the GstMemory objects were flagged with NO_SHARE.
This was because the buffer pool needed to be sure that when
a buffer returned to the pool, it would be safe to release the
OMX buffer back to OpenMAX.
With this change, this is no longer a restriction. What this
commit introduces is a new allocator that allows us to track
the GstMemory objects independently. Now, when a buffer returns
to the pool, it is not necessary for the memory to be released
as well. We simply track the memory's ref count in the allocator
and we return the OMX buffer back when the memory's ref count
drops to 0.
The reason for doing this is to allow implementing zero-copy
transfers in situations where we may need to copy or map a
certain region of the buffer. For instance, omxh264enc ! h264parse
should be possible to be zero-copy by using an OMX buffer pool
between them.
