GitLab

Change-Id: I75890c0f64f93f48299895d1e0bcfbf91846a4ab

The first level is turned on for speed 1.

Change-Id: I3dba0f0250b97a25e174cacc2a46ca7f76572c85

Removes features for now so that we only add features with very
small loss.

Change-Id: Ie50f6af2a6cc19dde5f682754a1f0adf4ec957a8

Change-Id: I4270d1260854ac27b68c5694ca8102b92bee6faa

Use a neural-network-based binary classifier to predict the first split
decision on the highest level of the TX size RD search tree. Depending
on how confident we are in the prediction we either keep full unmodified
TX size search or use the largest possible TX size and stop any further
search.

Average speed-up: 3-4%
Quality loss (lowres): 0.062%
Quality loss (midres): 0.018%

Change-Id: I64c0317db74cbeddfbdf772147c43e99e275891f

The "use_intrabc" flag is signaled with CDF. No need to keep a counter
for it.

Change-Id: Ia62ef8f264aa5ce2f6fceddc0b2a7d2032c73044

Disable it for sub8x8 4:1 partitions (4x16 and 16x4), because of
conflict with cfl.

Enable it for all the other sub8x8 partitions sizes.

BUG=aomedia:998

Change-Id: Ifdd907f0ac1f987981e81c166eb71978e6ea27c3

This experiment has been abandonned.

Change-Id: Ieabc6f365651e2d116a4505a3cc202add94d1386

Change-Id: Ifcc92df2e8c69752c1dbff2b447eb22035814389

We do this by upscaling the deblocked output as we save it into the
RestorationStripeBoundaries line buffers. (See save_boundary_lines in
restoration.c for the details)

The upscaling is done by calling av1_convolve_horiz_rs, which reads
off the edge of the frame and, of course, across tile boundaries. This
means we need to extend the frame borders before saving boundary
lines (hence the changes to decodeframe.c and encoder.c)

Change-Id: Ia096846898b20afe4737433d772f7277d4f71724

We reverted to using 3-tap filters. So 4-tap filters related code
will not be used any more.

Change-Id: I7f65cf227d2eb3e9785474e3b33d0bdbf489b1f1

This change does not alter the encoder/decoder behavior. Extra
precautions are taken to stop an attacker from exploiting these function
pointers.

Change-Id: I4e0704f016774f2d8fbaeb2a4caec12fc6e67ec1

Removed assertions were not properly set up and may cause decoding failure when USE_CASUAL_CTX is enabled. This CL does not change bitstream.

Change-Id: Ib9193cbda32f342335a79aca39e9cc49204a0ec9

Before this patch, striped loop restoration didn't restart correctly
on each tile row. Now, the loop restoration stripes start at the top
of a tile row in the same way as if it were the top of the entire
frame.

Change-Id: I0a88a28d7804b2f09d792ecbbf4f22f666f67012

Because we have an (effective) 3-pixel border around each
processing unit, and the local sums in the self-guided filter are
only taken over at most 5x5 regions, we have 1 pixel's worth of
spare border.

We can use this border to greatly simplify the filter: Instead
of calculating a 64x64 region of the A[] and B[] arrays, we can
calculate a 66x66 region. Then we don't have to deal with complicated
boundary conditions when generating the final 64x64 output block.

This also makes a few other related changes:
* The 'boxnum' function has been effectively redundant
  for a while - due to the way we do the 5x5 (or 3x3) windowing,
  the values we actually use are always (2r+1)^2. So we can skip
  calling this function if MAX_RADIUS <= 2

* We can remove the annoying special case for tiny processing units
  in the self-guided filter, as we no longer have to worry about
  border behaviour

* We change the SSE4.1 code to match the new C code, removing a ton
  of complexity. Further refactoring/speedups are probably
  now possible, but this includes the minimal changes to pass all
  the tests.

Change-Id: I99beee164a31349a5228a9bef048e5f35c9639f2

Change-Id: I299e2f2a1967f867a5452e0c449abe5243ac5d13

BUG=aomedia:994

Change-Id: I0fa9d48487256655798dbdd64acad523e84557c6

Change-Id: Ia5df312c759faa38fe336ab32a7d4908760ecf08

BUG=aomedia:995

Change-Id: Ic320e5de9b0c7d320b0f6dddce93f1445be61234

These are just RESTORATION_PROC_UNIT_SIZE shifted right by the
vertical or horizontal subsampling for this plane and it's easier not
to have to pass them around.

Change-Id: I86441d6cd86bb146f3e5dcdf2c89e34dd9fed0e1

Avoid switch fallthrough (and associated gcc/clang warnings),
reduce scopes and add consts.

Change-Id: I28d910d9d39ee8fe2c5618364af602af5be5c186

This experiment has been adopted, we can simplify the code
by dropping the associated preprocessor conditionals.

Change-Id: I02ed47186bbc32400ee9bfadda17659d859c0ef7

BUG=aomedia:992

Change-Id: Ifcaedaf312f056fcc29e6a8e020aac0ddc52affd

Change-Id: I6cf4e3bdc5eda5f7baabc7c6ad99ba49f69a8fa4

Change-Id: I3873acafcd9539da96f67328cdb8faf7798be90f

Pull the per-unit processing out into a couple of new functions,
to make the overall logic of search_selfguided_restoration()
a bit more obvious

Change-Id: Ib4ed9be7d4c76e22dc56f933f3f9d09160242f71

* The function av1_decode_tg_tiles_and_wrapup performs some per-frame
  initialization; some of this was mistakenly being performed
  once per tile group instead, leading to strange behaviour
  (eg, forgetting loop-restoration coefficients, forgetting
  the boundary information for all but the last tile group, etc.)

  Fix this by pulling all of the initialization code into its
  own function and calling it only if the initialize_flag is set.

* While fixing the above, I realized that the 'context_updated'
  flag in av1_decode_tg_tile_and_wrapup was not behaving as intended:
  The idea is that, when using frame parallel mode, we save the
  frame context early so that the next frame can start decoding.
  Then we don't need to store the frame context at the end of
  the frame, since we already dealt with it at the start of the frame.

  However, this 'context_updated' flag was local to one tile group,
  ie. it got reset to 0 once we started decoding the second tile group.
  So we'd end up storing the frame context again at the end of the frame
  if there was >1 tile group.

  This didn't break anything, but it is a bit weird. So, to match
  the original intent, we ditch the 'context_updated' flag and
  directly check if we're in frame parallel mode when necessary.

* Fix a bug where we read one byte too much from a tile group
  OBU when the extended OBU header was used.

BUG=aomedia:892

Change-Id: Ifbe561de0de35525d809e23915ac5263273e8de7

Previously we were calling aom_extend_frame_borders to generate
extended pixels for use in loop-restoration. This generates quite a
large border, when we only need 3 pixels.

In addition, we were also calling extend_frame, which does the same
thing but with a smaller border, once (in the decoder) or multiple
times (in the encoder) per plane.

This patch tidies all of this up so that we only call extend_frame
once per plane, with the largest border size we need (3px).

It also adds two new #defines. RESTORATION_BORDER is the 3 pixel
border needed to do filtering for a processing
unit. RESTORATION_CTX_VERT is the number of rows saved for each stripe
when doing striped loop restoration.

Change-Id: I2c3ffcc19808f79db195f76d857e2f23da5d8a84

After this patch, we don't scale sb coordinates vertically when using
HORZ_FRAME_SUPERRES.

Change-Id: I24c652b4b357b132e8b29979a119e7aeb8420e19

Before an increased border of 288 pixels was used when both
CONFIG_EXT_PARTITION and CONFIG_FRAME_SUPERRES were enabled.

However increased border is also required when only
CONFIG_EXT_PARTITION is enabled.

For example when:
1) current frame is 2x smaller than reference frames
2) block size is 128x128

BUG=aomedia:991

Change-Id: I09dfddfdf6bd0b0efde2556acb924bb563b6da2f

This is to ease integration with striped loop restoration

Change-Id: If10e16656a3fe42bcca3e7238e4e729c962f2bb8

BUG=aomedia:993

Change-Id: I555f430541413a42e9e14310bfde93304dc15cfa

This CL simplifies context derivation for nz and base level flags in
level map.
1. Reduce SIG_COEF_CONTEXTS from 58 to 42.
2. NZ and base level flags share the same context offsets derived from a
template of size 5 (down from 7).

In limited runs, compression performance seems neutral if not better.
Encoding time for a key frame on a local linux machine is reduced by about 25% or more.

Change-Id: Ibd93b21c839154bc5ae26b993f9e66537cbf5942

Change-Id: Ic5cf4c8393a00810700e42a162fbb36a62d9b56f

This name better reflects the functionality of the mode.

Change-Id: Idc2372e7c7b226f70f63fa4f272a88cb4d857089

With --enable-ext-partition, it is possible to have superblock size at
both 64x64 and 128x128. But MAX_SB_SIZE_LOG2 is defined to be 7, i.e.
128x128 size. In get_token_alloc(), the token count for a tile or a
frame is always computed based on "tokens/superblock * number of
superblocks", so it should use the actual superblock size, instead of
128x128.

The segfault in #988 is a result when every tile is defined at one
single superblock of 64x64, the token size computation for each tile
was assuming it is 128x128, resulting 4x overestimate of the token
count for every tile. This caused the pointer for many tiles to go
beyond the allocated token arrary, which is correctly allocated at
frame level.

BUG=aomedia:988

Change-Id: I2da21f181446ad7117ce3269f01a43950c379028

The speed feature simply restricts the number of depths
searched. Currently it is turned on by default for speeds>=1.
The coding efficiency impact (tested on lowres 30 frames) seems
to be ~0.15% and the speedup is in the order of 15%.

Change-Id: I514832bd7df937292875f73d9c9026e49ac576f2

Change-Id: I12471d18544a9634c28ed139b7211c9af0762927

USE_UNI_COMP_REFS is always true when ext-comp-refs is enabled and is
always false otherwise.

Change-Id: Ic6061a4ab88a2ebfeb11f178872ab13909060c0b

* add build-time option to add/read 4-byte size of every OBU
* update to R19 by adding forbidden bit to obu header
* add metadata OBU decoding

Change-Id: I525264f53a8666b7bc29493fb29ae661841eba50