GitLab

The experiment is fully adopted now.

Change-Id: I27906d2af4c746ce55aa17f64d1c0ef281e23ab2

Previously, segments using SEG_LVL_REF_FRAME were unable to signal the choices
of GOLDEN = 4, BWDREF = 5, ALTREF2 = 6 and ALTREF = 7, as
seg_feature_data_max[SEG_LVL_REF_FRAME] was set to 3. This patch increases the
value to 7 to account for these options.

BUG=aomedia:951

Change-Id: I9732fa2be96ead2d4b6efdbce34a92e43c7dd04e

Sequences starting with intra-only frames previously resulted in undefined
behaviour with CONFIG_AMVR == 1, as seq_force_integer_mv was only read for
keyframes.

This patch makes changes as follows:

- The syntax element force_screen_content_tools has been added to the
  SequenceHeader struct, and is read and written correspondingly

- seq_force_integer_mv has been renamed to force_integer_mv and moved to the
  SequenceHeader struct, and is read and written correspondingly (provided that
  force_screen_content_tools != 0)

- The conditional reading/writing of allow_screen_content_tools now happens for
  every frame after reading/writing error_resilient_mode (CONFIG_OBU == 1) or
  the sequence header (CONFIG_OBU == 0)

- The conditional reading/writing of cur_frame_force_integer_mv now happens for
  every frame after reading/writing allow_screen_content_tools

BUG=aomedia:1048

Change-Id: I689476fc2fa781dc8ec6fc8da91926cc8cfd3dc2

Removed redundancy code in ref_mv finding.

Change-Id: I4f9d0257a217e764cce820c01d66985778010de7

Clamp_mv_ref happens in multiple places in ref_mv search, which can be
convoluted as reported in issue 1124. This change is to consolidate
the clamping into one place.

Borg test result on lowres set:
avg_psnr:    ovr_psnr:   ssim:
  0.000       0.000      0.001

BUG=aomedia:1377
BUG=aomedia:1124
BUG=aomedia:857

Change-Id: I1649d5b5f37683c9c30e493c6eed13a808ab543a

When the coding block size has one side in length of 64 and above,
scale up the mfmv reference search step size from 8 to 16 along
that direction. The midres coding stats get 0.02% better. Among
all the finished hdres points, no negative results showed up.

BUG=aomedia:1379

Change-Id: I70ab7a9f9d1cf365d8ed1e06dbede307b6bc46ec

Reduce the ref mv search over spatial neighbors from 4 to 3
columns.

BUG=aomedia:1382

Change-Id: I44eb96e2ff4243d720a5f4f68be504995ebd69b6

Change-Id: I94f739840b90688506ebf5a4e7cf355cd6fb75ed

And some cosmetic changes.

Change-Id: Ibbc62bbb023d80ceee83120b4f6eb2a95b1d1a64

Compression changes are of noise level.

BUG=aomedia:1309

Change-Id: Icc0e607e2d16c9041a9b8bab1b267da21ebbddd0

Change-Id: Ife016b5adeeb26071fa20ce6f66b7b52074b36d6

Change-Id: I297b1572e89d5668529855f2436b47af95ea2f80

1) The scan orders of rect txfm will be diagonal with shorter side
first
2) Neighbor is not updated because lv_map doesn't use the neighbor
3) Write a scan test to protect the scan order.
4) Performance on key frame is neutral. Will report the overall
performance by comment later.

BUG=aomedia:1369

Change-Id: I417392f2d9870e035532a2e1cc7cf71dc20ea219

Change-Id: I22693eb22ab4c9cf15396cabbec226e186ad3a7f

Change-Id: I7cf9b4f82bce224cc327947779417dc91bf25184

Reorder intra mode coding so that the luma
direction is below luma intra mode and chroma
direction is below chroma intra mode.

BUG=aomedia:1357

Change-Id: Ie91d04abda7762e59e37e7a4831d206802fe7f2d

This is removed because it was decided to drop support for 4:4:0 Chroma
subsampling in AV1. High bit depth subsampling is added to rtcd so that
this solution is uniform for high and low bit depth.

Change-Id: I9abd6b34164069c0c66881e262ad60026694b948

This patch changes the bit stream slightly to make it impossible to code a loop
restoration unit size that is smaller than the superblock size.

BUG=aomedia:1291

Change-Id: I59637177ccb481d45c2108b0fea624a952e0cd03

Change-Id: Idcc1abaedd6615130fcf1741450fbc5d2f83705c

This is important because only one angle delta
is used for both chroma planes. Using PLANE_TYPE
increases readability and reduces the chance of
people trying to access the angle_delta for
AOM_PLANE_V (i.e. 2).

Change-Id: I26f452e6fcee8bd7fd1e670e73f457d6937e0da9

This will not cause coding stats change.

Change-Id: I0c527a8906f99170cf66d2360654bf65af81c7e0

The speed feature use_upsampled_ref was re-implemented in
https://aomedia-review.googlesource.com/c/aom/+/13647
and so its name was a misnomer.

Hence, change the name to a more appropriate one.

Change-Id: Id66fe76d6fcf720e98d01e4ea530ef52e79d063d

The behavior of this feature was recently changed to do
upscaling/convolve on the fly as needed -- instead of upsampling the
references. So, the name was a misnomer.

Also, as this doesn't require additional memory now, we can always use
it -- irrespective of --cpu_used / frame resolution.

Compression performance:
Overall: -0.15

Netflix_TunnelFlag:          -1.14
Netflix_Aerial:	             -0.66
touchdown_pass:              -0.49
Netflix_SquareAndTimelapse:  -0.40

https://arewecompressedyet.com/?job=with-tx64x64%402018-01-17T19%3A20%3A42.090Z&job=upsampled_ref_remove_limit%402018-01-18T19%3A36%3A53.025Z

Change-Id: Ic3d08fbff645121b2f95b279b8053cad130dab93

BUG=aomedia:1369

Change-Id: Ic3a6922e8a380883b429515ea5f73d31093385e1

Key-frame-only performance
lowres 0.084% gain
midres 0.14% gain
hdres 0.147% gain

BUG=aomedia:1369

Change-Id: Ic7e09583720730f81e87fc5f83b794cc5ce5fb98

The PSNR change is neutral.

BUG=aomedia:1369

Change-Id: Iade17a19580eed788338d5933423ea0235316952

Compression loss on screen_content testset keyframes: 0.3%

BUG=aomedia:1367

Change-Id: Ifb5dd0d5b6362ec3cb29f43bdb6a8af6f38f6fe5

Remove unneeded reference motion vector search function calls.
This does not change the coding stats.

Change-Id: Idbd43634f714e2f18d6e5f0a187721ed8ea35d52

As of commit 2d276b39, reconintra.c cannot compile with CfL disabled,
since the UV_* versions of the intra modes are not defined.

Fix this by moving the UV_PREDICTION_MODE enum outside of its
enclosing #if CONFIG_CFL block. This requires a couple of minor
adjustments elsewhere in the code.

Change-Id: Id8d10ebfa627d8528719b557933929719d863a87

read_palette_mode_info() shouldn't read luma palette info in monochrome mode.

BUG=aomedia:1311

Change-Id: I86f479beac41a3b8725926086f3d21605a003088

On the encoder side (encoder.c):
* seq_force_integer_mv is initialised to 2 (and never changed)
* if allow_screen_content_tools == 0, cur_frame_force_integer_mv is set to 0
* if allow_screen_content_tools == 1, then
  - if seq_force_integer_mv == 2, cur_frame_force_integer_mv = is_integer_mv(..)
  - otherwise cur_frame_force_integer_mv = seq_force_integer_mv

This behaviour is now reflected in the functions that read/write the
uncompressed header, with suitable asserts on the encoder side to ensure that
this is in fact what is occurring.

BUG=aomedia:1338

Change-Id: Ic508a9b2188002c6ea26501c142ed58cce98be6a

Use frame_offset to locate the location of mismatch.

Change-Id: I17ceb60febf5f30041a1d8e35556ae7034d095cd

Implement av1_lowbd_inv_txfm2d_add_4x4_sse2
Implement av1_lowbd_inv_txfm2d_add_4x8_sse2
Implement av1_lowbd_inv_txfm2d_add_8x4_sse2
Implement av1_lowbd_inv_txfm2d_add_4x16_sse2
Implement av1_lowbd_inv_txfm2d_add_16x4_sse2

A brief speed test shows that using the included SSE2 functions
completed by this CL, for speed1 lowbitdepth encoder speeds up >9%
and lowbitdepth decoder speeds up >25%, comparing to the highbitdepth
implementation in the baseline.

Change-Id: I0576a2a146c0b1a7b483c9d35c3d21d979e263cd

This patch uses the neighboring ref counts to design the contexts
for the coding of the first reference frame of a reference pair for
the compound prediction. This aligns the context design with that
for the second reference frame of a reference pair for the
compound prediction.

The new designed contexts are much simpler than that in the baseline.
The number of contexts for each binary symbol is reduced from 5 to
3. Further, the logic for each context only depends on the collected
neighboring ref counts, which is straightforward to derive.

The default CDFs for the first reference frame coding have been
updated using aom_entropy_optimizer.

Experimental results demonstrate a small coding gain for Google test
sets of both lowres and midres, with 30 frames coded for the default
coding tool setup:

lowres: avg_psnr -0.077%; ovr_psnr -0.076%; ssim -0.106%
midres: avg_psnr -0.059%; ovr_psnr -0.066%; ssim -0.037%

BUG=aomedia:1356

Change-Id: I781abbe4616dc3f3a7213ec663946ff9844eb830

Re-design the compound mode reference motion vector fetch. Use
a single run to provide all the compound ref mvs. Save the potential
additional two ref mv search runs on single reference frames.

Tested on night_720p 50 frames at 800 kbps. The average time cost
on find_mv_refs calls is reduced by 15% (average 69875 us ->
60473 us). The overall compression performance change is less than
0.01%.

BUG=aomedia:1373

Change-Id: I388b9cf36817d10613cd2c9d0bd8865b43324009

Continued to refactor the reference MV search code, so that we can
combine the compound ref mode search later.

Change-Id: I6227a3142e82caa20f2a17a0c76147035eaa2129

This commit re-designs the reference motion vector fetching from
spatial neighbors with different reference frames from a current
coding block. Instead of re-running through the VP9 like reference
motion vector search, it goes through the nearest top row and left
column only. Such process kicks in, if and only if the regular
reference frame match based mv search didn't find 2 or more mvs.
The search through neighboring blocks with different reference
frame will stop once 2 mvs are found.

To decide the reference mvs, it compares the reference frame types
from the two blocks. If they are on the same side, directly re-use
it. Otherwise, reverse the sign of the motion vector.

The compression performance change is in the noise range as
0.03% down.

BUG=aomedia:1372

Change-Id: Ib698d7c463f2f42c767f6ca008c8a7c84289df60

Implement av1_lowbd_inv_txfm2d_add_32x64_sse2
Implement av1_lowbd_inv_txfm2d_add_64x32_sse2
Implement av1_lowbd_inv_txfm2d_add_16x64_sse2
Implement av1_lowbd_inv_txfm2d_add_64x16_sse2

Change-Id: I1b27618f153583cc787e7bf6ef1616e7c6932990

Implement av1_lowbd_inv_txfm2d_add_8x32_sse2
Implement av1_lowbd_inv_txfm2d_add_32x8_sse2

Change-Id: Ibd5de72e1d2c4dabba5af020a06e8cfac329dc3d

1. Reorder functions to align with TX_SIZE define order.
2. Merge functions for each TX_SIZE which have very similar code
   into a universal function lowbd_inv_txfm2d_add_internal_sse2.
3. No speed impact was spotted except size 8x8, so the 8x8 version
   stays unchanged.

Change-Id: Ic896aacd93745906716582af855774807a863231