GitLab

Reduces precision in find_affine_int() function. Makes the maximum
mv allowed 512 from 1024.
Negligible impact on coding efficiency.

Change-Id: I76d4c6824528e3f940d1275fe0bd22d71015a8d0

Change-Id: I4b77388fa49e29a8302fe8c3276471e8156ce2d5

This will make the code cleaner

Change-Id: I94622b696b0a8d51d28dbbe29c767a62942bc5f3

This was introduced in 72e2e982

Change-Id: Ia4d0a0606a518964d5f6f257874f8f3d5b4d651b

high-latency, cpu=0:
  PSNR | PSNR Cb | PSNR Cr | PSNR HVS |   SSIM | MS SSIM | CIEDE 2000
0.0817 | -0.4672 | -0.6651 |   0.1043 | 0.0968 |  0.1063 |    -0.1847

low-latency, cpu=0:
  PSNR | PSNR Cb | PSNR Cr | PSNR HVS |   SSIM | MS SSIM | CIEDE 2000
0.0860 | -0.7110 | -0.7393 |   0.1018 | 0.0946 |  0.1800 |    -0.1537

low-latency, cpu=4:
  PSNR | PSNR Cb | PSNR Cr | PSNR HVS |   SSIM | MS SSIM | CIEDE 2000
0.0097 | -0.8022 | -0.7514 |  -0.0115 | 0.0621 | -0.0092 |    -0.3804

Change-Id: I5c2eced295598771c7a74ec15a5f0eb84d872cc9

BUG=aomedia:395

Change-Id: I381f16e66e2540c9bf24727abf8915a3850dcc92

Change-Id: Id32a6a21ffda48ef16b027b55f9082614f389003

Running this test in Jenkins causes all but trivial changesets to
fail unless the hash is changed. Other changes in master
can meanwhile cause the hash to be out of date by the time
of merge even if it is changed, so rebases of previously
passing changesets will then fail again.

Change-Id: I4ee5010ceec85da7eedd8ada2d171f64a92491c1

This commit fixes a residual parameter set conflicts in function
copy_dering_16bit_to_16bit(), due to the change of cb4x4 support
approach in cdef to better respect od_dering.c as a codec
independent file.

Change-Id: I7e2295cf6a083729077374b13836909bb611d573

Change-Id: Ifd55343b701eb753537b9711abd377adcedde8ec

Change-Id: Ibe19f4422324680b0d3132eab79bdc035911d79d

Change-Id: I7ca73e03ed171b57a05dd1fd5957906e11b66728

Change-Id: I0f4576522a07597dbb04089b02ca1fae67075ba4

Move most of the code of inverse_transform_block to
av1_inverse_transform_block such that encoder can use the
function as well.

Use av1_inverse_transform_block in av1_encoder_block_intra.
This will make the code cleaner.

Change-Id: I84dbeef2c65e252910606dbea446ce43165d504c

This will simplify the code flow in av1_subtract_txb()

Change-Id: I575da4a74f435b5862d2a848bdedae9856672121

A simpler interface for intra block prediction

Change-Id: I97e3f385746928d87999e6862b1d71a197df0302

This function help us merge subtract block code in
av1_subtract_plane and av1_encode_block_intra.

Change-Id: Ie793d88a218f1082c6fe28900a521f461e34d564

Change-Id: I463d01ecf0ef2d666c38f9dec510b8b0b1eecadc

It's less invasive and helps make future changes easier.

Change-Id: I0f858462f14109ace4d76a31daeac4289c94bc6e

EXT_TILE: allocate variables on the heap due to the potentially
large number of tiles.

ANS: initialise tile contexts.

EXT_INTER, MOTION_VAR, WARPED_MOTION: fix compilation.

Change-Id: I43a9d01c2442d97b6a8875712f77d932dce3740d

A bug, but no change to BDR.

Change-Id: I17fbe836fd72d148a4f0608fcff7e2fbcaf75eb0

Change-Id: Idae4934424b3e0c2de659e3f55ea6a47b5b4ceab

Change-Id: I6173dfb181777343d254aef880a8c9d379206556

On x86 there is no direct v128_shr_n_s8 equivalent, so
v128_cmplt_s8(a, v128_zero()) is much better than v128_shr_n_s8(a, 7).
Should have no impact on NEON.

Also replace v256_from_v128(v128_from_v64(a, b), v128_from_v64(c, d))
with v256_from_v64(a, b, c, d).

Change-Id: I711e3cb250689089d7b5336a294e9d6bdd998445

Previous code was using options selected for N=8 rather than the chosen
number of options.

low-latency, cpu=4:

  PSNR | PSNR Cb | PSNR Cr | PSNR HVS |    SSIM | MS SSIM | CIEDE 2000
0.0077 |  0.0286 | -0.1166 |  -0.0059 | -0.0479 |  0.0025 |    -0.0101

Properly save the best strengths

Change-Id: I629e929c2bc7a0a9592a9e49bfd7898d95174235

Change-Id: I46101bf22f38a642850b875fb333e7c6b9c1d5b8

Add boundary testing for both dering and CLPF.  Also add a CDEF prefix
to make it easy to select all CDEF tests.

Change-Id: Icd4fa4d778889cff1fd4a4e01c2b901703a431d3

Change-Id: I416ab6976702f7840f7ef1d86f737d51fa07ebbc

Change-Id: I1320b9e2f57363686340f066e5366be385daa493

Change-Id: I5243432106c2456f1220adb9d8f24ae5e4249748

Change-Id: I529509e4e997ba123799a3a581d20624d75cf582

This makes it clear when profiling that the correct SIMD optimised
function is run.

Change-Id: I35d69b3611f40650a85f1973c4010453b2bf5a53

Rather than having the adjustment in the leaf functions, do the
adjustment in the top-level function.  Optimising compilers would
figure this out themselves as far as the functions are inlined, but
probably no further and this patch gives a slightly reduced object
code size.

Change-Id: I104750962f613fa665391c9b2a9e99bcc6f47f93

As part of 9cf0c9cd the buffering was made
to better handle 4:2:2, but that causes regressions in the tests, so we're
backing out part of it for now.

Change-Id: I9ca4cfeb159aa65514613989e3dcbc30f86ec5b2

Only 1 sample needs to be collected. Max of 8 neighbors are
used.
In LS estimation, the projection samples (sx, sy)->(dx, dy) are
intentionally smoothed by assuming 3 shifted versions
(sx, sy+n)->(dx, dy+n), (sx+n, sy)->(dx+n, dy), (sx+n,
sy+n)->(dx+n, dy+n) also contribute to the estimation.
For example, instead of using A[0] = sx^2, we use the sum of
squares of source x of four points, A[0] += 4sx^2+4*n*sx+n^2.
But computational cost wise, it does not add much overhead. Coding
gain is mostly same as the old version. If no smoothing is added,
will lose 0.3% on lowres.

Change-Id: I04be32cffa525f7dc8ee583c0bf211d7bdc6e609

Loss of gain in AWCY
HL 0.23%
LL 0 (since no compound is used in LL)
lowres 0.277%
midres 0.248%

Change-Id: I46ad1e2f07411c838f2ca6765de57a60a9c68b12

Take all sub8x8 neighbors as 8x8 blocks and use mv assigned to
the last block.

Change of performance in AWCY
HL improved by 0.01%
LL improved by 0.06%

Change-Id: I55d3c5401222396d871f9157b62b3de29e5390b0

Change-Id: I5b202d8e57dbc8fc283f2fda7afe0fec0c3ef622

Change-Id: I392b801f61b0d3bcd1cd6157ab783f76ea8c9e5e

Change-Id: I60543913cbd8dc5cad524ab74697227f9e93836e