Rewriting GstCudaConverter object, since the old implementation was not
well organized and it's hard to add new features.
Moreover, the conversion operations were not very optimized.
Major change of this implementation:
* Remove redundant intermediate conversion operations such as
any RGB -> ARGB(64) conversion or any YUV -> Y444 (or 16bits Y444).
That's not required most of cases. The only required case is
converting 24bits (such as RGB/BGR) packed format to 32bits format
because CUDA texture object does not support sampling 24bits format
* Use normalized sample fetching (i.e., [0, 1] range float value)
and also normalized coordinates system for CUDA texture.
It's consistent with the other graphics APIs such as Direct3D
and OpenGL, that makes sampling operations much easier.
* Support a kind of viewport and adopt math for colorspace conversion
from GstD3D11 implementation
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3389>
GstCudaConverter object can do colorspace conversion and scale at once.
Adding new element "cudaconvertscale" to do that, this can
save unnecessary GPU operation if colorspace conversion and
rescale is required for given input stream format.
Most of codes are taken from d3d11convert element
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3389>
Introduce a new API that can return a GstTypeFind * with helper functions
and data set around buffer data.
While at it, drop factory field from GstTypeFindBufHelper. While it was
useful for logging, it was not passed through function arguments and keeping
it for logging would require an additional API increasing the API surface
and making it harder to use.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3296>
Instead of returning a "const gchar" or a "gchar" that should not be freed, always
return a duplicated string as those functions were used together with g_strdup anyway.
This is needed to prepare support for returning modified strings in next commit.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1147>
This is small regression from commit f7abd81a.
When calling `gst_element_query()` no pad is associated with that query, but the
current code always forwards the query to the associated pad, which is NULL in
previous case. This patch checks for the pad before forwarding the query.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3404>
If we don't receive any data from usrsctp, then there will be no src pad
for the stream id and the stream reset will fail to remove the relevant
src pad. Workaround by first attempting to add the relevant src pad, then
almost immediately removing it.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3381>
Replace video_copy with memcpy to fix the issue when the sizes of the
src frame and dst frame don't match. Moreover, for Windows, you have to
do the copy first and call gst_msdk_import_to_msdk_surface later.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3231>
Replace video_copy with memcpy to fix the issue when the sizes of the
src frame and dst frame don't match. Moreover, for Windows, you have to
do the copy first and call gst_msdk_import_to_msdk_surface later.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3231>
Currently MSDK context does not support d3d11va. Now introduce d3d11va
device to MSDK context, making it able to create msdk session with d3d11
device and to easily share with upstream and donwstream.
Add environment variable to enable user to choose GPU device in multi-GPU
environment. This variable is only valid when there's no context
returned by upstream or downstream. Otherwise it will use the device
that created by upstream or downstream.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3231>
Add support for more formats so as to run the libvpx high bit depth test suite.
This means the files under CONFIG_VP9_HIGHBITDEPTH
This also allows running the yuv444p 8bit file in the regular 8 bit vp9 suite.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3356>
If an input is malformed (only produces cea608 field 1 cc_data) then
when in passthrough we would effectively be dropping every second cea608
on output as we would not store any unused cea608 data.
Fix by having all code paths go through the framerate conversion code
which will store and retrieve any relevant data across buffers.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3211>
When a tile format is padded and imported as DMABuf, the stride
contains the information about the actual width and height in
number of tiles. This information is needed by the detiling shader
in order accuratly calculate the location of pixels. To fix that,
we also copy the offset and strides into the otuput format and
the converter will ensure that the shader is recompiled whenever
the stride changes.
This fixes video corruptions observed when decoding on MT8195
with videos that aren't not aligned to 64bytes in width.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3365>
... otherwise PAR can be wrongly signalled during the negotiation
Fixing below pipeline when desktop resolution is not 640x480
gst-launch-1.0.exe \
d3d11screencapturesrc ! videoscale !
video/x-raw,width=640,height=480,pixel-aspect-ratio=1/1 ! d3d11videosink
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3360>
1. Removes the verification if the internal encoder is not opened
yet to allow the property setting.
2. And toggles on the base class' reconf flag for each property
variable that can be modified at run time.
3. Mark those modifiable properties as mutable while playing.
Currently the run-time modifiable properties are:
qpi, qpp, qpb, bitrate, target percentage, target usage and rate control
Other properties can be enabled too, but they need testing.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2466>
Adds an internal function reset() which drains the internal queues and
calls the reconfig() vmethod.
This reset() method is called inconditionally at set_format() and in
handle_frame() if the instance's reconf flag is enabled.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2466>
If parameters remain similar enough to avoid either encoder reopening
or downstream renegotiation, avoid it.
This is going to be useful for dynamic parameters setting.
To check if the stream parameters changed, so the internal encoder has
to be closed and opened again, are required two steps:
1. If input caps, profile, chroma or rate control mode have changed.
2. If any of the calculated variables and element properties have
changed.
Later on, only if the output caps also changed, the pipeline
is renegotiated.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2466>
This method will return the caps configured in the reconstruct buffer
pool, and its maxium number of buffers to allocate.
The caps are needed later to know if the internal encoder has to be
reopened if the stream properties change.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2466>
This adds a new boolean property `auto-reconnect`, defaulting to `true`.
Setting it to `false` makes the elements (in caller mode) immediately
report an error to the application instead of trying to reconnect.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3326>
Adding DirecShow video capture filter mode, in addition
to existing MediaFoundation and WinRT(UWP) mode, to support
DirectShow only filters (not KS driver compatible)
such as custom virtual camera filters.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3350>
Use gst_debug_set_threshold_from_string's new reset behavior to undo
GST_DEBUG and ensure the logging tests have a known configuration.
`gst_debug_set_threshold_from_string ("LOG", TRUE)` has the same effect
as `gst_debug_set_threshold_from_string ("", TRUE)` followed by
`gst_debug_set_default_threshold (GST_LEVEL_LOG)`.
Don't bother remembering the default log level set when the test
started. It will get reset by the next test, anyway.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/605>
TLDR: Make `gst_set_threshold_from_string ("", TRUE)` reset *all*
threshold settings, including those set by previous invocations of
`gst_debug_set_threshold_from_string`.
The docs say:
@reset: %TRUE to clear all previously-set debug levels before setting
new thresholds
What actually happens is it sets the default threshold to `ERROR`,
leaves the patterns in place and calls
`gst_debug_category_reset_threshold` on each category.
In effect, any category that is matched by a pattern gets reset to that
threshold if the app changed it by directly invoking
`gst_debug_category_set_threshold`. All other categories are reset to
`ERROR`.
In my opinion this parameter currently has little value, as the same
effect can be achieved by including `ERROR` (without a pattern) in the
string, as in `"foo*:WARNING,*bar:INFO,ERROR"`.
What I actually expect it to do is reset *all* threshold settings,
including those set by previous invocations of
`gst_debug_set_threshold_from_string`, starting off with a clean slate
for the patterns provided with the call.
Otherwise there is no API to do this, besides:
- Painfully removing patterns one-by-one via
`gst_debug_unset_threshold_for_name` *if* you know what the patterns
are.
- Adding a `*:FOO` pattern to affect all categories, which makes the
default threshold useless and practically leaks all the old
patterns.
In my opinion this also makes it fit better into the layers of threshold
config, which is:
1. Temporary:
- `gst_debug_category_set_threshold`
- `gst_debug_category_reset_threshold`
2. Patterns:
- `gst_debug_set_threshold_for_name`
- `gst_debug_unset_threshold_for_name`
- `gst_debug_set_threshold_from_string`
- `GST_DEBUG`
3. Default:
- `gst_debug_set_default_threshold`
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/605>
The scenario is the following:
* Thread 1 is pushing an EOS event on a sinkpad
* Thread 2 is pushing a STREAM_START event on the same sinkpad before Thread 1
returns. Note : It starts pushing the event after Thread 1 took the object lock.
There is a potential race between:
* The moment Thread 1 sets the EOS flag once it has finished sending the
event (via store_sticky_event). When it does that it has both the STREAM and
OBJECT lock
* The moment Thread 2 sends the STREAM_START event (Which should release that
EOS status), but removing the EOS flag is only done while holding the OBJECT
lock and not the STREAM_LOCK, which means it could be re-set by Thread 1 before
it then checks again the EOS flag (without the STREAM lock taken).
The EOS flag unsetting by STREAM_START should be done with the STREAM lock
taken, otherwise it will be racy.
Fixes https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/1452
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3320>
In a few cases throughout qtdemux, the results of QT_UINT32 were being
stored in a signed integer, which could cause subtle bugs in the case of
an integer overflow, even allowing the the result to equal a negative
number!
This patch prevents this by simply storing the results of this function
call properly in an unsigned integer type. Additionally, we fix up the
length checking with stsd parsing to prevent cases of child atoms
exceeding their parent atom sizes.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3344>
Subtracting a gint from another (or a guint from another) has no guarantees that
it will result in a gint.
Therefore do the actual comparision instead.
Also use the *actual* type for comparing flags (the field value types are different)
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3319>
Unlike the legacy elements, GstAdaptiveDemuxStream is a GObject now,
so a bunch of things that were actually stream methods on the
parent demux object can directly become stream methods now.
Move the stream class out to a header of its own.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3314>
Sometimes g_input_stream_read_all_finish() can return
0 bytes, but still succeed (return TRUE) and have more
data available later. Only finish the transfer
if it returns 0 bytes *and* FALSE with no error.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3314>
The cancelled flag was only set in the stream finalize()
method, after all activity on the stream has stopped anyway.
Replace uses of cancelled with checks on the stream state.
Remove the replaced flag, which was checked but never set
to TRUE anywhere any more.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3314>
With non-serialized sticky events, such as GST_EVENT_INSTANT_RATE, we both want
to store the event (for later re-linking) *AND* push the event in a non-blocking
way.
We therefore must *not* propagate pending sticky events if the event is "sticky
or serialized" but only if it's "serialized"
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3254>
- Make the srt_epoll_wait loops more uniform.
- Error only via GError when possible; let the element send the error
message. Avoids a second error message.
- Return 0 when cancelled. Avoids an error message from the element.
- Don't send an error message from send_headers when we're a server
sink.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3156>
When using the child proxy notation (child::property=value) it may
happen that the target child does not exist at the time of parsing
(i.e: decodebin creates the encoder according to the contents of the
stream). On this cases, we want to delay the setting of the property
to later, when new elements are added. Previous logic performed a
delayed set even if the target child was found but the property
was not found in it. This should be treated as a failure because,
unlike missing elements, properties should not appear dynamically.
By not failing, typos in property names may go unnoticed to the end
user.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2908>
These null checkes are slightly misleading when double-checking
mutability for external language interop. None of the functions in
these files allow the variable at hand to become `NULL` under normal
operation, because they are checked at initialization and never (allowed
to be) reassigned to `NULL`.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1615>
When matching segments across playlists with Program-Date-Times,
use the difference in segment PDTs to adjust the stream time
that's being transferred. This can fix cases where the
segment boundaries don't align across different streams
and the first download gets thrown away once the PTS
is seen and found not to match.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3309>
Check whether the init file / MAP data for a segment
is different to the current data and trigger an
update if so. Previously, the header would only
be checked in HLS after switching bitrate or
after a seek / first download.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3307>
Previously the minimum buffering threshold was hardcoded to a specific
value (10s). This is suboptimal this an actual value will depend on the actual
stream being played.
This commit sets the low watermark threshold in time to 0, which is an automatic
mode. Subclasses can provide a stream `recommended_buffering_threshold` when
update_stream_info() is called.
Currently implemented for HLS, where we recommended 1.5 average segment
duration. This will result in buffering being at 100% when the 2nd segment has
been downloaded (minus a bit already being consumed downstream)
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3240>
This is an additional quality parameter. In the default configuration this
quality switch is deactivated because it would cause a workload increase
which might be significant. If workload is not an issue in the application
it can be recommended to activate this feature.
A flush request is done when set_format is called to empty internal bit
buffer maintained by fdk-aac. When this happens, during the explicit
call to handle_buffer, decodeFrame does not return a AAC_DEC_OK. This
gets reported as a decoding error while no decoding error in fact took
place. Since this can be confusing, just return a GST_FLOW_OK and log
that an explicit flush was requested.
In order to figure out if the "raw" audio contained within the wav
container is actually DTS, right now we call the typefinder helper
which runs all typefinders.
Speed up this type finding process by specifying the extension.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3294>
GST_TRACERS="leaks" GST_DEBUG="GST_TRACER:7,leaks:6" gst-play-1.0 --use-playbin3 test.mkv
When running a pipeline like above, leaks are observed.
0:00:56.882419132 240637 0x5562c528ccc0 TRACE GST_TRACER :0:: object-alive, type-name=(string)GstConcatPad, address=(gpointer)0x7efd7c0d20a0, description=(string)<'':sink_0>, ref-count=(uint)1, trace=(string);
0:00:56.882429131 240637 0x5562c528ccc0 TRACE GST_TRACER :0:: object-alive, type-name=(string)GstConcatPad, address=(gpointer)0x7efd7c0d2be0, description=(string)<'':sink_0>, ref-count=(uint)1, trace=(string);
0:00:56.882437056 240637 0x5562c528ccc0 TRACE GST_TRACER :0:: object-alive, type-name=(string)GstConcatPad, address=(gpointer)0x7efd7c0d3720, description=(string)<'':sink_0>, ref-count=(uint)1, trace=(string);
gst_element_release_request_pad does not unref the pad. It needs to
be followed by gst_object_unref. Doing that fixes the above leaks.
Use g_ptr_array_new_with_free_func with gst_object_unref as the free
function to unref the pad after release.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3177>
Posting latency messages causes a full and potentially expensive latency
recalculation of the pipeline. While subclasses should check whether the latency
really changed or not before calling this function, we ensure that we do not
post such messages if it didn't change.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3282>
The picture parameter picture->top_field_first is reused in this mode
to signal the TOP fields. As a side effect, it will change every frame
and current code assumed that if this changes then a renegotiation is
needed. Fixed this by ignoring that change whenever we are decoding one field
only.
Fixes#1523
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3276>
In fact, all the h264 bit writer have byte aligned output except
the slice header. So we change the API from bit size in unit to
byte size, which is easy to use. For slice header, we add a extra
"trail_bits_num" to return the unaligned bits number.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3193>
We use va pool as msdkvpp's bufferpool, which means both va memory
and dma memory will be allocated by va pool. Considering drm modifier
stuff is not ready, we use va memory with higher priortiry than
dma memory when deciding vpp caps.
Besides, this patch removes the specified "interlace-mode" in vpp caps.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3253>
