INTER_REFS_PER_FRAME specifies the maximum number of reference frames
that each Inter frame may use.
TOTAL_REFS_PER_FRAME is equal to INTER_REFS_PER_FRAME + 1, which
counts the INTRA_FRAME.
Further, at the encoder side, since REF_FRAMES specifies the maximum
number of the reference frames that the encoder may store, REF_FRAMES
is usually larger than INTER_REFS_PER_FRAME. For example, in the
ext-refs experiment, REF_FRAMES == 8, which allows the encoder to
store maximum 8 reference frames in the buffer, but
INTER_REFS_PER_FRAME equals to 6, which allows each Inter frame may
use up to 6 frames out of the 8 buffered frames as its references.
Hence, in order to explore the possibility to store more reference
frames in future patches, we modified a couple of array sizes to
accomodate the case that the number of buffered reference frames is
not always equal to the number of the references that are being used
by each Inter frame.
Sarah Parker [Tue, 2 Aug 2016 19:28:41 +0000 (12:28 -0700)]
Adjust gm parameter computation to avoid mismatch
The gm parameters need to have WARPED_PRECISION_BITS precision
until they are written to the bitstream because functions in
reconinter use these parameters before they are written to
the bitstream. Previously, the parameters weren't being converted
to WARPED_PRECISION_BITS until they were read from the bitstream
which causes an encode/decode mismatch.
Yue Chen [Tue, 2 Aug 2016 22:18:44 +0000 (15:18 -0700)]
Fix a bug and a function name in EXT_INTRA experiment
(1) Apply ALLOW_FILTER_INTRA_MODES flag to the correct place, otherwise
there are bitstream mismatchs when it is 0.
(2) Rename pick_ext_intra_iframe() to pick_ext_intra_interframe().
Manually cherry-picked the following changes: 8c8d16de vp9 -> vpx in names 75b57d39 VP9_ -> VPX_ in function names 761a7088 VP9_INTERP_EXTEND -> VPX_INTERP_EXTEND 4273a52c VP9->VPX in border pixel macros 03568c31 VP9_FRAME_MARKER -> VPX_FRAME_MARKER 2334f51d VP9->VPX in fdct function names
Jingning Han [Wed, 27 Jul 2016 16:12:53 +0000 (09:12 -0700)]
Separate frame context index for different frame types
This commit makes the encoder to use different frame context index
for different frame types. In the baseline setting, it sets the
frame context index of the overlay frame to be different from other
regular inter frames. In the ext-refs setting, it further allows
the backward reference frame to use a different index.
It improves the compression performance for both settings.
Ivan Krasin [Mon, 25 Jul 2016 19:11:02 +0000 (12:11 -0700)]
Fix compilation error under Clang 4.0.
The LLVM trunk has reached 4.0 and now __clang_major__ is not enough
to distinguish between old XCode Clang and the new 'real' Clang.
Using __apple_build_version__ allows to make this distinction.
Added a new expt rect-tx to be used in conjunction with ext-tx.
[rect-tx is a temporary config flag and will eventually be
merged into ext-tx once it works correctly with all other
experiments].
Added 4x8 and 8x4 tranforms for use initially with rectangular
sub8x8 y blocks as part of this experiment.
There is about a -0.2% BDRATE improvement on lowres, others pending.
When var-tx is on rectangular transforms are currently not used.
That will be enabled in a subsequent patch.
Yi Luo [Fri, 15 Jul 2016 23:41:30 +0000 (16:41 -0700)]
HBD fast path quantization speed improvement
- HBD encoder speed improvement (SSE4.1):
Enable CONFIG_VP9_HIGHBITDEPTH, on Xeon E5-2680,
50 frames, park_joy_1080p, 12-bit,
Encoding time reduces from 4846481 to 4177471 (ms)
- Add unit test to verify bit-exact and EOB calculation
Sarah Parker [Mon, 11 Jul 2016 18:47:55 +0000 (11:47 -0700)]
Add global motion parameters to compressed header
Currently nothing is implemented to compute GM parameters, this
just adds the capability to send them in the bitstream if they
were computed. Still need to implement the reconstruction
based on the parameters in reconinter.
Wei-ting Lin [Tue, 12 Jul 2016 20:19:45 +0000 (13:19 -0700)]
Allow OVERLAY frames to use the show_exsiting_frame flag
ARF with zero strength temporal filter can be reused by setting the
show_existing_frame = 1, and in this case, there is no need to
refresh the reference frame buffer. However, we used the flag
"refresh_golden_frame" as the identifier for the starting point of a gf
group.
A new flags "is_arf_filter_off" is used to record if the filter with
strengrh zero is used.
Johann [Wed, 22 Jun 2016 23:08:10 +0000 (16:08 -0700)]
Merge changes from libvpx/master by cherry-pick
This commit bring all up-to-date changes from master that are
applicable to nextgenv2. Due to the remove VP10 code in master,
we had to cherry pick the following commits to get those changes:
Add default flags for arm64/armv8 builds
Allows building simple targets with sane default flags.
For example, using the Android arm64 toolchain from the NDK:
https://developer.android.com/ndk/guides/standalone_toolchain.html
./build/tools/make-standalone-toolchain.sh --arch=arm64 \
--platform=android-24 --install-dir=/tmp/arm64
CROSS=/tmp/arm64/bin/aarch64-linux-android- \
~/libvpx/configure --target=arm64-linux-gcc --disable-multithread
Use quotes whenever possible and {} always for variables.
Replace multiple set_all calls with *able_feature().
Conflicts:
build/make/configure.sh
vp9-svc: Remove some unneeded code/comment.
datarate_test,DatarateTestLarge: normalize bits type
quiets a msvc warning:
conversion from 'const int64_t' to 'size_t', possible loss of data
mips added p6600 cpu support
Removed -funroll-loops
psnr.c: use int64_t for sum of differences
Since the values can be negative.
*.asm: normalize label format
add a trailing ':', though it's optional with the tools we support, it's
more common to use it to mark a label. this also quiets the
orphan-labels warning with nasm/yasm.
CONVERT_TO_BYTEPTR(x) was corrected in: 003a9d2 Port metric computation changes from nextgenv2
to use the more common (x) within the expansion. offsets should occur
after converting the pointer to the desired type.
Speed test shows the new vertical filters have degradation on Celeron
Chromebook. Added "X86_SUBPIX_VFILTER_PREFER_SLOW_CELERON" to control
the vertical filters activated code. Now just simply active the code
without degradation on Celeron. Later there should be 2 set of vertical
filters ssse3 functions, and let jump table to choose based on CPU type.
improve vpx_filter_block1d* based on replace paddsw+psrlw to pmulhrsw
Make set_reference control API work in VP9
Moved the API patch from NextGenv2. An example was included.
To try it, for example, run the following command:
$ examples/vpx_cx_set_ref vp9 352 288 in.yuv out.ivf 4 30
Jingning Han [Mon, 18 Jul 2016 17:14:13 +0000 (10:14 -0700)]
Align the quantizers for inter/inter modes in the first pass coding
Use regular extended zero bin quantizer for both inter and intra
modes in the first pass. This doesn't affect lowres and midres
significantly, but would bring back 0.9% coding gains for hdres.
projects / libvpx / log