In surface YUV output mode the width/height fields of the VpxOutputBuffer were
never populated. Fix this by adding a new method to set the width/height and
calling it from JNI like we do for GL YUV mode.
PiperOrigin-RevId: 250449734
Update configuration script to use an external build, so we can remove use of
isysroot which is broken in the latest NDK r19c. Also switch from gnustl_static
to c++_static so that ndk-build with NDK r19c succeeds.
Issue: #5922
PiperOrigin-RevId: 250287551
This is part of the public API and the underscored versions
weren't intended to be used outside the library.
Also updating the external instructions to checkout the version
of libvpx that supports this macro for this particular constant.
#codehealth
PiperOrigin-RevId: 242170926
There were some edge cases in which we'd forget to release DRM
sessions. For example if we read a format and acquired a
pendingDrmSession (in onInputFormatChanged), then immediately
read another format and overwrote pendingDrmSession, we'd
forget to release the one that's been overwritten.
This change hopefully makes release much clearer. We keep a list
of all drm sessions we're currently holding. Whenever we update
either drmSession or pendingDrmSession, we release any other
sessions that are in the list.
PiperOrigin-RevId: 228905465
Turn on flag to enable fast loop filtering. This improves
multi-threaded decode performance by starting loop filtering early
if there are free threads available to do so.
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=218239442
Renamed to stop collision between the ExoV1 and ExoV2 implementation.
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=217348150
With default of value set to -1, every single dropped frame is reported because of expression: if (droppedFrames >= maxDroppedFramesToNotify) { maybeNotifyDroppedFrames(); }
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=213502573
Not having this annotation may cause Kotlin implementations to fail.
Issue:#4802
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=212980643
Power Comparison:
TextureView 1080p 720p
H264 HW 498, 496 507, 478
VP9 RGB 1050, 1104 1185, 1152
VP9 ANativeWindow 1070, 985 700, 674
GLSurfaceView
VP9 YUV 1075, 1112 716, 635
SurfaceView
H264 HW 419, 409 397, 377
VP9 RGB 1044, 1139 654, 671
VP9 ANativeWindow 975, 835 617, 623
VP9 MediaCodec 683, 679 488, 476
Measures average current drawn mAH on a Nexus 6 at full brightness from time t=3 to t=95 seconds. The same clip was used for all tests. Two measurements were taken for each category.
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=203847744
Video renderers assume that the player position is advancing linearly while in
the started state. MediaCodecVideoRenderer schedules frames for rendering in the
future in the expectation that the player position is advancing.
This assumption is not currently true when using a position from the AudioTrack:
the player position can be fixed for (in the worst case) up to about 100 ms
before it starts increasing. This leads to an effect where the first frame is
rendered then a few other frames are rendered, then there's a pause before
frames start being rendered smoothly.
Work around this issue by checking whether the position has started advancing
before scheduling frames to be rendered in the future.
It might be preferable to make the audio renderer only become ready when its
timestamp can start advancing, but this is likely to be complicated.
Issue: #3841
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=190937429
Also remove logic for tracking the next output buffer in LibvpxVideoRenderer, as
this allowed many consecutive frames to be rendered that were actually late
after dropping to keyframe. It looks better to show frames at a consistent
100 ms rate.
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=188144739
The pending exception will be thrown upon returning to Java
from native, but we should return early rather than continuing
to execute the native method to the end so as to avoid undefined
behavior. Note that the return value is irrelevant (because the
pending exception will be thrown).
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=184119606
Try to delay failure for as long as possible. That is, propagate
DRM session failures only after an encrypted buffer arrives or clear
sample playback without session is not allowed.
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=183076348
Make LibvpxVideoRenderer non-final and add protected methods to match
MediaCodecVideoRenderer.
Reorganize methods to separate BaseRenderer, protected and internal methods.
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=181320714
Now this counter includes input buffers too, which are dropped as part of
skipping to keyframes for catch up.
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=171119930
In out libvpx extension, currently we always call vpxGetFrame and copy the data
from the native decoder to output frame. However, if the inputBuffer has
isDecoderOnly set, we can avoid populating the output buffer, but only setting
BUFFER_FLAG_DECODE_ONLY.
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=170318527
Using the same sampler introduced some minor horizontal scratches.
-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=167347995
