BUG=aomedia:693

Change-Id: I09b34abc41ee6f85619f5a05f668e06491e542f0

The logic that searches for motion vectors in scan_row_mbmi and
scan_col_mbmi steps in 4x4 units when the block size is less than
8x8. That test (bsize < BLOCK_8X8) doesn't work when you have 16X4 or
4X16 blocks, because they appear higher up in the enum.

Change-Id: Idb1a97f8a43b675fd78dbc76ae501fcdff7adbcd

Angie Chiang authored 7 years ago and James Zern committed 7 years ago

Turn off the assertion because av1_init_lv_map doesn't pass the
count according to the position into get_base_ctx_from_count_mag()

The speeding-up optimize_txb is still on-going.
I will turn the assertion on once the speeding-up process is
getting to a milestone.

BUG=aomedia:704

Change-Id: I0119bf825bedc1fccbe4963f341a17f0e2601d02

This will allow the mrc-tx experiment to pass in its own color map
and set of probabilities to the tokenize and bit packing functions.
The corresponding change on the decoder side will come in a followup
shortly.

No change in performance.

Change-Id: I1872b7f1b1cf5b102e6289388ce1ae449e67cb4d

There are six pieces of code in reconinter.c and two in rdopt.c which
iterate over the blocks along the top or left edge of the current
block for OBMC. Before this patch, each bit of code has its own
implementation of the iteration, which is reasonably finicky to get
right.

This patch factors out that logic into a pair of helpers
(foreach_overlappable_nb_above and foreach_overlappable_nb_left). The
functions take a "fun" parameter, which contains the loop body. Note
that the iteration is too complicated for us to be able to define a
macro that could be used like

  FOREACH_NB_ABOVE(rel_pos, nb_size, nb_mi) { ... }

While C's syntax doesn't seem to let you do that, once the compiler's
optimisation pass is done inlining everything, the results are
essentially the same.

The iteration logic is also slightly generalised: the old code checked
whether a block was shorter or narrower than 8 pixels by comparing a
block size with BLOCK_8X8. This doesn't work when you have a 4x16 or
16x4 block because e.g. BLOCK_16X4 is not less than BLOCK_8X8. This
generalisation is (unsurprisingly) needed in order to to support 16x4
or 4x16 blocks.

This patch doesn't address the CONFIG_NCOBMC functions in reconinter.c
that do prediction from right and bottom edges.

This patch shouldn't affect the generated bitstream in any way: the
code is supposed to be equivalent.

Change-Id: I9e5a116b012c18645604a7d98fb98be99697d363

Make it explicit that:
- forward tx is using the matrix
- inverse tx is using the transposed matrix

Change-Id: Ia9507f4f68397b07b6fd1868bb421bb9d5dd817d

The selfguided filter code was sometimes fetching 8 bytes of data
when it only needed 4. This was normally fine, but lead to problems
in the selfguided filter test when compiling for x86-32, where we
accidentally read off the end of the input buffer.

Fix this by only reading the amount of data we actually need.

BUG=aomedia:700

Change-Id: I2448b7b0d9cb2f9292a092675a66da64c89f913c

This reverts commit 38cdce8e.

Reason for revert: <INSERT REASONING HERE>

It is reported that this introduced a segment fault, revert now to investigate. 

Change-Id: I229dcdc09462ae4eb9cf2a3da4ce59aa9eafe65e

Change-Id: I5da893be07b3b9aa4351b5bb1af987f4992370ef

This is undefined behaviour in C and might confuse the optimizer, leading to incorrect code.

Change-Id: Ia4bb60478068da678f013bdd6ab6a49814d89ebe

Improves reconstruction performance.

Change-Id: Ia30a345e1f5d4e450138c1dd3a7b700119c7de31

Change get_lgt in order to integrate a later experiment
lgt_from_pred with lgt. There are two main changes.

The main purpose for this change is to unify get_fwd_lgt and
get_inv_lgt functions into a get_lgt function so the lgt basis
functions can always be selected through the same function in
both forward and inverse transform paths. The structure of those
functions will also be consistent with the get_lgt_from_pred
functions that will be added in the lgt-from-pred experiment.

These changes have no impact on the bitstream.

Change-Id: Ifd3dfc1a9e1a250495830ddbf42c201e80aa913e

Make the lv-map rate-distortion optimization account for the
per symbol cdf update. This improves the level map coding
performance by 0.4%.

Change-Id: I1cbecafbad4e6642abc3782c9741636d92d87eb7

Change-Id: Icdce2ac2203bacdcbbbf17d1fc3308dc1ed15974

Change-Id: Ife4e7e4f62b217fd55db1ec3ef69c243f1651777

Allow per symbol cdf probability model update in the lv-map
encoding process.

Change-Id: Id0c4a659c04ffa981d8f04537eda2f1304d1682a

Support the per symbol cdf update at the level map syntax element
decoding process.

Change-Id: Id58bf86fb1c7c72dd14ddaa3781caaa27f446ef7

Initialize the cdf model for level map syntax elements.

Change-Id: I3865e07c126eb4c856803c12485b05782dea6526

When frame-superres is used with loop-restoration use a 4-tap
upscaler to reduce complexity.

Change-Id: Idf5712b9a006763c4000e264176b04aa8331b12f

Change-Id: Iec280ba1e314849ac027085f61e62a48dd224be4

Change-Id: I1b789acc18f1e69fb5db069ccd8bd17815938e9d

Change-Id: I3cc9cb8135ec5e8aa20c8cff901a2f37b3b459ee

BUG=aomedia:685

Change-Id: I0698b281fbaff2e77a50f5dc9a10d752de6c6f79

Change-Id: Idcc4c4722bd0379c8667b60115bfd85cf779a6ac

Doing this means that we don't have to memset temporary buffers to
zero in reconinter.c, which was taking ~5% of cycles in a short
encoding test (using perf to attach to a running encode).

Change-Id: Ibb6e31920000b876c6ee99f454d89c8a97e9fb31

Change-Id: I48ebf352c6c28e5c0c0e477b24828f0e3fe1dedb

Change-Id: Ie7c216218bd233e74970b261186df8f08aca6193

This will speed up lv_map by 4.3%

Change-Id: I886f67b9f1ae2e5f567e114fa34cf3ba437b8381

Change-Id: I6e9e1c94f688ed84c49d3b861dbac15c37248827

For sub-8x8 blocks, scan loop should have at least one iteration.
Previously, loop would be skipped if col_offset == 1 and n8_w == 1.

This change appears to have neutral effect on bdrate (+0.01).  This could
mean that extra MV candidates are not helping that much for sub-8x8.

Change-Id: Ic778312b39113ab8f348f5c7c18813593a831b7c

The orders_32x128 and orders_128x32 variables are only used when
CONFIG_EXT_PARTITION is false. Only define them in that case to avoid
a compiler warning.

Change-Id: Iebe297eff36b139caa83d039101cf16eb88ca8f8

The change is purely an encoder-side tuning, which constrains a GF
group interval to be either 8, or 12, or a number larger than 12
(maximum of 16). The change is under ext_refs.

This work has been part of the summer internship effort by chendixi@.

Further, this CL removes the experiment of flex_refs and merges it
into ext_refs, as flex_refs has been completely focused on the encoder
side optimization for ext_refs.

On Google test sets, above (stillness + interval enforcement) achieve
a coding gain in BDRate as follows, compared against the AV1 baseline
with all tools enabled by default (excluding ext-tx, global-motion,
and convolve-round for speed concern):

lowres: avg_psnr -0.193%; ovr_psnr -0.370%; ssim -0.332%
midres: avg_psnr -0.556%; ovr_psnr -0.680%; ssim -0.798%

Change-Id: Ia992d8ef0d82cc0ef2caba5eec9b26f0358e6bde

This will speed up lv_map exp by 8%

Change-Id: Ideaebfd855a3b2e7c2a07ac028407e7336b86697

This will speed up lv_map encoder time by 1.5%

Change-Id: Ia9f7634bae221d2fb8fb7dc6f3202c6bacdd5b32

the coverage flags are required at link time as well as compile time

Change-Id: Ic581f8ce747615f1fa1dc13d989d6e233e982572

CONFIG_UV_LVL --> CONFIG_LOOPFILTER_LEVEL
Change name because the experiment combines filter level search
in U, V plane, and horizontal, vertical search in Y plane.

Change-Id: Ia05b8f80cd93b60a0305eb94e5783559ef904c13

A new experiment for deblocking filter that separates vertical
and horizontal filter strengths. This experiment is based on the
assumption that non-flatness characteristics of vertical and
horizontal direction may differ. Therefore selecting different
filter strengths for vertical and horizontal can improve deblocking
performance.

The process of finding proper filter strength:
1. Search through the filter level under the constraint that
   (vertical == horizontal), and find the best solution.
2. Fix vertical level as the best solution found in step 1 and vary
   horizontal level to find the best value.
3. Fix the selected horizontal level, vary vertical level to find
   its best value.

The experiment is working with UV_LVL, sharing the same config flag.
The searching for horizontal and vertical filter strength only applies
on Y plane for now.

The experimental flag should be changed to filter_level later.

Change-Id: I164eec8d3ccb3da7ff109c5c55f4b52c1536ddf1

Currently nothing forbids wedge from being signalled when
the block is > 32X32, even though there is no corresponding wedge
mask for that block size.

BUG=aomedia:640
BUG=aomedia:636

Change-Id: I538be0229a12b5ef01b2e5a950c9f16ef9a5c51e

BUG=aomedia:697

Change-Id: Id7e971e515c9faa3df32e162befd57907278549d

This commit resolves an encoding failure issue due to the
tokenization in palette when the lv-map experiment is turned on.

BUG=aomedia:689

Change-Id: I8e476ae81e4d5d11c6aab7611dff9deb7fa63e9e