GitLab

reset_skip_context() was always clearing the entropy contexts for
all three color planes, using a block size that corresponded with
the luma plane.

However, when chroma_2x2 is disabled, then for sub-8x8 luma block
sizes, the corresponding chroma block size is always 4x4, and the
skip flag only affects the chroma blocks corresponding to the
upper-left luma block.

This patch makes reset_skip_context() reset the contexts that
actually correspond to the chroma blocks that are skipped (if any).
It also moves reset_skip_context() to av1_reset_skip_context() in
blockd.c, because blockd.h gets included before onyx_int.h, which
declares the required is_chroma_reference() function.
reset_skip_context() was too large and used in too many places to
be a reasonable candidate for inlining, anyway.

AWCY results on objective-1-fast:

cb4x4-fix-base@2017-05-11T06:26:50.159Z -> cb4x4-fix-reset_skip@2017-05-11T06:28:45.482Z
  PSNR | PSNR Cb | PSNR Cr | PSNR HVS |   SSIM | MS SSIM | CIEDE 2000
0.0301 |  0.1068 |  0.1463 |   0.0359 | 0.0260 |  0.0347 |     0.0479

A regression (near the noise range), but without this fix, the line
buffer size required by the entropy contexts will be doubled.

Change-Id: I12fa6e60d9c1c7c85927742775a346ea22b3193f

Create named constants for the minimum TX_SIZE used in the luma
plane, as well as the minimum allowed TX_SIZE for which we have to
explicitly code a selection (i.e., one larger than the minimum luma
TX_SIZE).

Then, use these constants consistently with tx_size_to_depth() to
index the tx_size counts in FRAME_COUNTS, and also consistently use
TX_SIZE named constants to index the tx_size_implied counts.

Failing to index these counts correctly can, e.g., cause tx_mode to
be chosen incorrectly.

Change-Id: I706a62a33e2282e67c97a68bade87fb8023ec13b

Clear the clang warnings when var-tx is enabled.

Change-Id: I6b005124159b0607ec3284cc2e5ba287745a2541

BUG=aomedia:545

Change-Id: I8beacce0f87b52dfd614bcd8c88426839c4ba51d

Unifies the chroma and luma paths for estimation of the
restoration parameters.

No change in bit-stream or results.

Change-Id: I9998c269cc00ded8b94d0e660b5441a4f9b94848

Use recursive transform block coding order.

Change-Id: I6d4671fe669e8a1897e034973de181078272cbfd

Encode a block line by line, horizontally or vertically. In the vertical
mode, each row is predicted by the reconsturcted row above;
in the horizontal mode, each column is predicted by the reconstructed
column to the left.

The DPCM modes are enabled automatically for blocks with horizontal or
vertical prediction mode, and 1D transform types (ext-tx).

Change-Id: I133ab6b537fa24a6e314ee1ef1d2fe9bd9d56c13

cb4x4 adds a new TX_2X2 to the TX_SIZES enum, which offsets the
indexing of blockzero_probs. Add a set of default probabilities for
this case (simply copied from the TX_4X4 probabilities).

These will only be used if chroma_2x2 is enabled (though moving the
other probabilities back to their correct offset will affect all
other sizes).

This is an encoder-only change.

AWCY results on objective-1-fast:

cb4x4-fix-reset_skip@2017-05-11T06:28:45.482Z -> cb4x4-fix-blockzero_probs@2017-05-11T06:29:26.242Z
   PSNR | PSNR Cb | PSNR Cr | PSNR HVS |    SSIM | MS SSIM | CIEDE 2000
-0.1520 | -0.2682 | -0.0972 |  -0.2342 | -0.1638 | -0.1701 |    -0.3278

Change-Id: I5594f19af2935c8a8d43105a27ad33f6222ed7f2

Change-Id: Ia0eb3a3694fdbe9d33548ff4014b704b2f3db86a

BUG=aomedia:526

Change-Id: I8271a3c18ac9fa53b6304eb5e2a07ba787b05724

Always use translational motion predictor for sub8x8 blocks.

BUG=aomedia:552

Change-Id: I82bbf6c9d2a2e91783774966336f5e2bc87c3d13

Since chroma_sub8x8 is a subset of chroma_2x2, both of these
experiments should not be turned on at the same time.

Change-Id: If54e0a3005d6fe26ade274700678558baf977f9c

Before this commit set_frame_size decided on the frame size to use and
then set it. It has been changed to just take an arbitrary size to make
it more flexible.

The decision of what size to use has been moved to the new
set_scaled_size and existing av1_calculate_superres_size.

setup_frame_size implements the old logic of getting the scaled size,
then superres size, then setting the frame to use it.

This change will make post-encode upscaling possible using
set_frame_size since it can now be given a target resolution.

Change-Id: I2f0f20740a54f751bc9963f279d6bcfdfb2722c2

Change-Id: I6d3cdcc3ee39e82edb201833da20b7412cc29038

When the above and left neighbor blocks are coded in the sub8x8
block size, use 4x4 level prediction block to produce the reference
for obmc.

Change-Id: Ic19f6e3b3880cebd153078823a3b77b04e28053b

Exposed by Change-Id: I048c6e9cc790520247cc21ae9b92a9c8d84d00a7

BUG=aomedia:525

Change-Id: Ia83f8a8efcf0eac4912f247f38887c0dd533da85

Fix a bug introduced in the previous commit that effectively
disables chroma-sub8x8.

Change-Id: I521fdc1f15f5bed5272081217e70d1e2549c7ecd

This is similar to the ones for 32x32, 32x16 etc.

Change-Id: Iec97f4dc2b0109d39a14e9d8e2d5241a79b9e529

The superres scale was not initialized correctly, which prevented
regular resizing if CONFIG_FRAME_SUPERRES was set. This patch
initializes superres scaling to 1:1 so it won't interfere with anything
else.

Change-Id: I9e3c89e91a4eb66146b9b924e8c19d0ed44a8dc4

Frame level guided filters are now used as an option for
chroma components in addition to the Wiener filter.

Change-Id: Ie33299d22c15b69741ede55686177b7b8ce8e2b3

Use a single struct for tokenization and level map coding.

Change-Id: Id685992b7db5964ee204c4b0b90379df50c56546

Properly scale the chroma block size accroding to the YUV format.

BUG=aomedia:522

Change-Id: I2c42763394eea6383f81443bf1098fe05f448bd5

Change-Id: Id6078c5e3c82bc382af7b576133beb9026797a9b

Change-Id: Ib0bb35ff31bf317bb7eb09a1738649137f791589

* set dependent tile to be default
* make dependent tile to work with supertx

Change-Id: I2bcc73c6e6df46ca0c2fbdae7c4f2944738fff79

Use a variant of joint_motion_search to improve the motion search
for compound blocks (both average and masked types) with only one
NEWMV component.

Change-Id: I7cae812dced24acde638ae49869e6986557ce7dd

This adds the proper cfgs to av1_{inv/fwd}_txfm1d_cfg for
the identity transform so all hbd transforms can use
the same codepath. This has no impact on performance
since the new identity transforms that correspond with
the cfgs are not yet being called. Once this is checked in,
we should be able to delete all deprecated transform functions
and have a single code flow for all hbd transforms.

BUG=aomedia:524

Change-Id: I3d1bfbc8bc29b367e8ddf7dcd27525af0bd31067

Change-Id: I33e78f2677083df7b7385f4a2989e9d599bbf5c7

The re-designed sub8x8 coding block system settles. As part of the
research attempt, the sub8x8-mc experiment can now be removed.

BUG=aomedia:477

Change-Id: Icb1bf52471c560c1357e9121fbc5aa78ed854421

Change-Id: I1939cfe48570d9c797fd29d12ebcc9213388951a

av1_set_size_literal took unsigned ints as its size inputs. This was
inconsistent with elsewhere in the code base so this change makes it
take signed ints instead.

Change-Id: I5f39f813714f5d6cb1fc93f928f33b3633b653ec

Limit the minimum prediction block size for chroma component to
be 4x4 in OBMC. This resolves an enc/dec mismatch when all the
experiments are turned on.

BUG=aomedia:548

Change-Id: I3151912387943b03ef5c9aa2d10fe13d15c79cf7

It gives 0.1% gain on lowres and midres

Change-Id: I555a492a68571c525713840d73aa5614fe80a87d

Add functions which take both components of a masked compound and
compute the resulting SAD/SSE. Extend joint_motion_search to understand
masked compounds, and use it to evaluate NEW_NEWMV modes.

Change-Id: I782199a20d119a6c61c6567df157508125ac7ce7

-9.1 BD-RATE PSNR on wikipedia_420 1st keyframe
-1.04 BD-RATE PSNR on twitch-1 set 1st keyframe

Change-Id: Ia9cc5b26bc587bcc2375d137768ac3af3238fe0c

Improved the performance a little bit compared to using the full set

Change-Id: I28a96459fa77d7e1d5c101d162faa2effd525341

The hbd transform configurations were originally written for all possible
2d transforms. Now that there are many more possible 2d transforms
due to EXT_TX and RECT_TX, it is simpler to write the cfg for the
4 1D transform types and compose them to make all new possible transform
types. This will allow for an easier integration of the identity transform
for EXT_TX and rectangular transforms for RECT_TX into the current
hbd transform codepath and facilitate the removal of obsolete transforms.
This has no impact on performance.

BUG=aomedia:524

Change-Id: I1e217bcd217fd637b1df94fae62d9c59a0523c1a

Change-Id: I640cf08ac583ff5092a18e81cd0e33f05609db6b

Fixes a baseline BDRATE drop issue introduced in
0743857c.

Partially reverts the above patch.

BUG=aomedia:544

Change-Id: Ia7cd89330cab189d22b58e183951fc6429021f42

Subset1:
  PSNR | PSNR Cb | PSNR Cr | PSNR HVS |   SSIM | MS SSIM | CIEDE 2000
0.0000 | -0.0001 | -0.0000 |  -0.0001 | 0.0001 |  0.0000 |    -0.0001
https://beta.arewecompressedyet.com/?job=cfl%402017-05-18T16%3A08%3A54.160Z&job=cfl_fix%402017-05-18T16%3A08%3A02.454Z

Change-Id: I6e403b8a0a14803d38db881031ad51878c683d65