This fixes a memory leak. When dropframe-threshold has been set,
libvpx may output less frames than the input ones, which causes
some GstVideoCodecFrames to queue up in GstVideoEncoder's internal
frame queue with no chance of ever being all released. And because
the frames keep references to the input buffers, the input buffer
pool keeps allocating new buffers and memory usage grows very fast.
For example the following pipeline's memory usage grows at a rate
of about 1GB per minute!
videotestsrc ! capsfilter caps=video/x-raw,width=1920,height=1080,framerate=30/1,format=I420 ! \
vp8enc target-bitrate=1000000 end-usage=cbr dropframe-threshold=95 ! fakesink
https://bugzilla.gnome.org/show_bug.cgi?id=783086
https://github.com/mesonbuild/meson
With contributions from:
Tim-Philipp Müller <tim@centricular.com>
Jussi Pakkanen <jpakkane@gmail.com> (original port)
Highlights of the features provided are:
* Faster builds on Linux (~40-50% faster)
* The ability to build with MSVC on Windows
* Generate Visual Studio project files
* Generate XCode project files
* Much faster builds on Windows (on-par with Linux)
* Seriously fast configure and building on embedded
... and many more. For more details see:
http://blog.nirbheek.in/2016/05/gstreamer-and-meson-new-hope.htmlhttp://blog.nirbheek.in/2016/07/building-and-developing-gstreamer-using.html
Building with Meson should work on both Linux and Windows, but may
need a few more tweaks on other operating systems.
When a frame's duration is too low, calling gst_util_uint64_scale()
to scale its value can result into it being truncated to zero, which
will cause the vpx encoder to return an VPX_CODEC_INVALID_PARAM error
when trying to encode.
To prevent this from happening, we simply ignore the duration when
encoding if it becomes zero after scaling, logging a warning message.
https://bugzilla.gnome.org/show_bug.cgi?id=765391
All code paths for handle_frame() must somehow take ownership of the frame, be
it by actually unreffing, forwarding the frame elsewhere or storing it for
later.
http://bugzilla.gnome.org/show_bug.cgi?id=760666
With the VPX decoders it's not simple to use downstream buffer pool,
because we don't know the image size and alignment when buffers get
allocated. We can though use GstAllocator (for downstream, or the system
allocator) to avoid a copy before pushing if downstream supports
GstVideoMeta. This would still cause a copy for sink that requires
specialized memory and does not have a GstAllocator for that, though
it will greatly improve performance for sink like glimagesink and
cluttersink. To avoid allocating for every buffer, we also use a
internal buffer pool.
https://bugzilla.gnome.org/show_bug.cgi?id=745372
This adds an automatic mode to the threads property of vpxdec in order to
use as many threads as there is CPU on the platform. This brings back
GStreamer VPX decoding performance closer to what is achieved by other
players, including Chromium.
https://bugzilla.gnome.org/show_bug.cgi?id=758195
Some files may provide different caps insight of one stream. Since
vp9enc support caps reinit, we should support cache reinit too.
If more then file cache file will be created, the naming will be:
cache cache.1 cache.2 ...
Based on patch by: Oleksij Rempel <linux@rempel-privat.de>
https://bugzilla.gnome.org/show_bug.cgi?id=747728
Some files may provide different caps insight of one stream. Since vp8enc
support caps reinit, we should support cache reinit too.
If more then file cache file will be created, the naming will be:
cache
cache.1
cache.2
...
https://bugzilla.gnome.org/show_bug.cgi?id=747728
The framerate very often is just an indication of the ideal framerate, not the
actual framerate of the stream. By just using the framerate, we confuse the
rate control algorithm algorithm as multiple frames will map to the same PTS
or have durations of 0.
https://bugzilla.gnome.org/show_bug.cgi?id=749122
... instead of just counting frames. The values are supposed to be in timebase
units, not frame units. This fixes various quality problems with VP8/VP9
encoding and in general makes the encoder behave better.
Thanks to Nirbheek Chauhan for noticing this bug.
