DRAFT: Rebase av2enc on main - #5331

…OMediaCodec#5083)

Initialize `ref_costs_single` for INTRA and TIP frame indices to
`INT_MAX` when they are not allowed.

Previously, these elements were left uninitialized in
`estimate_ref_frame_costs` if the respective modes were not allowed, but
they were still being read and passed to other functions in
`av2_rd_pick_inter_mode_sb`, triggering MSAN warnings.

Fixes AOMediaCodec#5081

Adds Section 6.9.1 conformance check.

Fixes AOMediaCodec#5085

…89) (AOMediaCodec#5087)

Fixes AOMediaCodec#1477

The default value for this feature is 1, fix the comment.

prune_split_with_ml/prune_none_with_ml should be reset to 0. Set
prune_split_ml_level/prune_split_ml_level_inter to - 1.

When the encoder is run in error-resilient mode
(`--global-error-resilient=1`), switch frames are emitted without their
reference frames being marked as restricted. This produces a bitstream
that is not AV2-compliant and that decoders reject.
  
This MR sets `restricted_prediction_switch` for error-resilient switch
frames as well, so switch frames generated in that mode carry the
correct restricted-reference semantics.

In error-resilient mode (`--global-error-resilient=1`), the current AVM
encoder converts alt-ref–type frames (which are reordered/future frames)
into switch frames. When such a switch is left unconstrained
(non-restricted), the encoder can assign it a low display order hint
(e.g. 1) while a frame with a higher display order hint (e.g. 4) has
already been output, producing a non-monotonic output order. Because a
non-restricted switch does not reset the OrderHint counter, this
violates the conformance constraint below (AV2 spec §7.3.6, "Coded
extended layer unit"):

``
If monotonic_output_order_flag is equal to 0, it is a requirement of
bitstream conformance that within a coded video sequence, for a given
value of obu_xlayer_id and obu_mlayer_id, if a coded output frame unit X
has an associated OrderHint value equal to ohX, there shall not be a
coded output frame unit Y in the same extended layer and embedded layer
that appears later than X in output order and has an associated
OrderHint value less than or equal to ohX, unless a switch frame with
restricted_prediction_switch equal to 1 appears between X and Y in
coding order.
``
``
NOTE: The value of OrderHint is reset at the start of a new coded video
sequence and at a switch frame with restricted_prediction_switch equal
to 1. In both cases, the OrderHint counter is effectively restarted,
allowing OrderHint values to be reused in subsequent coded output frame
units.
``

…iaCodec#5097)

The `ref_frame_sign_bias` array in the `AV2Common` structure is
populated in `av2_setup_frame_sign_bias()` but never used. This patch
cleans up the unused array and the associated code.

For tests with dynamic layer dropping:
CDF_AVERAGING should be set to 0/off, to avoid DPB desync.

Also for the MultiLayerTest2Embedded2TemporaSframe test
enable_explicit_ref_frame_map_ should be set to true.

Failures were seen at the lower speed settings (e.g., speed 2), not at
the speed 5 setting used in the test.

In the reference commit, `cm->ref_frame_side[]` and
`cm->ref_frame_relative_dist[]` were computed in two separate paths:
either inside `av2_setup_motion_field()` when `allow_ref_frame_mvs` is
true, or in `av2_setup_ref_frame_sides()`. In this patch,
`av2_setup_ref_frame_sides()` is called unconditionally to populate
these reference frame arrays, thus removing the duplicate computations
inside `av2_setup_motion_field()`.

(1) Add checkout step + common setup.
This resolves the issue where the 'Compare hashes' step fails with
'fatal: not a git repository' because the git command was run without
checking out the repository.

(2) Use `github.event.pull_request.head.sha` to get base commit SHA
instead of `github.sha` which is the SHA for the "temporary merge
commit". This ensures that the `git log` command to find commits
containing `STATS_CHANGED` keyword works correctly.

Fixes AOMediaCodec#5106

In the av2_check_profile_interop_conformance() function, the local
variable `profile` is of the BITSTREAM_PROFILE type, so it is redundant
to cast `profile` to BITSTREAM_PROFILE. I suspect the intended cast is
`int`, to match the %d printf format specifier.

Alternatively, we can simply remove the BITSTREAM_PROFILE casts.

…diaCodec#5068) (AOMediaCodec#5069)

This fixes app-only builds where ENABLE_APPS=ON but ENABLE_TESTS,
ENABLE_EXAMPLES, and ENABLE_TOOLS are OFF. avmdec/avmenc still need
libyuv/webm include paths and object sources.

Verified locally:
```
cmake --build build_fast --target avmdec -j8
cmake --build build_fast --target avmenc -j8
```

Fixes AOMediaCodec#5068

Update enable_explicit_ref_frame_map_ setting, and remove duplicate
tests (since we should set enable_explicit_ref_frame_map_ = true for
these tests).

Failures were seen at the lower speed settings (e.g., speed 2), not at
the speed 5 setting used in the test.

As extended sdp is allowed only for partition type less than uneven 4
way search partition, this change has no effect on stats. This is done
for consistency.

This change unifies uneven 4way partition search functions and h
partition search functions into one function

No stats changed.

- Search across ext partitions is rolled into a loop

No stats changed

- Moved pruning using ml results logic to a function
- Removed duplicate/unused initializations
- Handled a case where none partition search gets skipped if
  search_none_after_rect is false and search_none_after_split is true
  and search none is delayed to after rect based on split results
- Moved terminate partition search after none search into none search
  function so that it gets called correctly after none evaluation
- Restored rdmult original in an early return case

The code under CONFIG_COLLECT_PARTITION_DATA is stale. This commit
restores its functionality.

Check the requirement of bitstream conformance that SeenFrameHeader is
not equal to is_first_tile_group. If this requirement is not met, the
decoder may skip the read_uncompressed_header() call in
av2_read_tilegroup_header(), which eventually leads to a null pointer
dereference.

Fixes AOMediaCodec#5142.

The void cast is used to ignore the tp_orig variable in Release builds
because tp_orig is only used in an assert(). So the void cast should be
applied to the variable itself rather than the expression *tp_orig.

This is a port from libaom:
 fd74e18c14 Add the void cast to the tp_orig variable itself

Co-authored-by: Wan-Teh Chang <wtc@google.com>

…5175)

avmdec and avmenc are no longer added to AVM_EXAMPLE_TARGETS, so we
don't need to filter out avmdec and avmenc in ${AVM_EXAMPLE_TARGETS}.

The read_tile_syntax_info() and read_sequence_tile_info() functions are
defined and only used in av2/decoder/decodeframe.c.

av2_highbd_warp_affine and av2_ext_highbd_warp_affine are called
frequently
during inter prediction at encoder speed 1. Adds NEON intrinsics for
both
functions, achieving 3-5x speedup over C at the kernel level.

Micro-benchmark results (Apple M2 Ultra):

| Kernel | C (ns) | NEON (ns) | Speedup |
|--------|--------|-----------|---------|
| warp 32x32 | 9209 | 1741 | 5.3x |
| warp 16x16 | 2323 | 483 | 4.8x |
| warp 8x8 | 589 | 153 | 3.8x |
| ext_warp 16x16 | 1851 | 441 | 4.2x |
| ext_warp 8x8 | 453 | 158 | 2.9x |

CTC Results (RA, cpu-used=1, 33 frames, classes A4+A5):

| Metric | Delta |
|--------|-------|
| Encode time | -5.5% |
| BD-rate Y | 0.000% |
| BD-rate Cb | 0.000% |
| BD-rate Cr | 0.000% |

Unit tests updated.

Some av2_get_chroma_format_idc() calls in av2/common/level.c and
av2/common/timing.c can be omitted because the callers don't use the
output argument chroma_format_idc.

After these unneeded av2_get_chroma_format_idc() calls are removed,
the subsampling_x, subsampling_y, monochrome parameters of some
functions become unused and need to be removed.

In AV2, monochrome requires subsampling_x == 1 && subsampling_y == 1.
(This somewhat arbitrary constraint comes from AV1.)

Also check the return value of av2_get_chroma_format_idc() at a few call
sites.

This PR enables the 12-bit profile by default and brings its
implementation into conformance with AV2 Spec Draft with 12-bit support:
AOMediaCodec/av2-spec-internal#776.

1. Enable CONFIG_TESTONLY_12BIT_SUPPORT by default
- Flips the 12-bit config default from 0 to 1 so 12-bit encode/decode is
built in by default. (Superseded by the rename below; the flag is now
CONFIG_12BIT_PROFILE.)

2. Align MAIN_4xx_12_IP2 profile with AV2 spec Annex A
- Renames CONFIG_TESTONLY_12BIT_SUPPORT to CONFIG_12BIT_PROFILE, and the
profile enum TEST_ONLY_12BIT_PROFILE to Main_444C_12_IP2 (spec
Main_444C_12_IP2, seq_profile_idc = 5).
- Accepts 8/10/12-bit and all chroma formats (4:0:0/4:2:0/4:2:2/4:4:4)
for profile 5, per Table A.1 (previously 8/10-bit were rejected).
- Sets the max embedded-layer count for profile 5 to 3, matching its
IOP2 limit (Table A.3).
- Gives profile 5 its own PicSize/Bitrate factor row (36 / 3.0) per
Table A.2.
- Updates the in-code Annex A profile/config tables (A.1/A.2/A.5/A.6)
and comments, including the bit_depth_idc mapping.

Fixing issue AOMediaCodec#5200

This patch refactors and abstracts different pruning logics
within the intra mode evaluation loop into a new function and
also avoids redundant computations in intra mode loop.

This is a bit-exact change.

Abstracted code related to multi winner mode processing
into a helper function `refine_winner_intra_mode_tx`.
Also replaces memcpy with `av2_copy_array()`.

This patch renames variables in av2_rd_pick_intra_sby_mode()
for better readablity and code consistency.

- Updated comments to adhere to coding guidelines.
- Used local variables to avoid repetitive access from
data structures.
- Modified code logic for better readability.

Abstracted code related to luma intra mode cost
calculation into a helper function to improve
readability.

Coding stats shall remain the same.

The speed feature wienerns_refine_iters was in speed 2.

Enabling it in speed 1 shows speedup with minor quality loss.

Testset          PSNR-YUV   EncTime
A1 (17 frames)    0.02%      95.58%
A2 (33 frames)    0.03%      99.08%

STATS_CHANGED

Change-Id: I91fd3184f45190347c19b49eaf7bf1934238e5ea

… speed 4 (AOMediaCodec#5294)

Skip unit size search for Loop Restoration based on pyramid level for
speed >= 4.

Anchor: e96e164 
Tests: RA, A1 17 frames, others 33 frames

cpu-used=4:
+----------+--------+-------+--------+--------+-------------+----------+
| Summary | Y | U | V | YUV | Enc-time | Dec-time |
+----------+--------+-------+--------+--------+-------------+----------+
| A1 | 0.00% | -0.02% | 0.12% | 0.00% | 95.621% | 99% |
| A2 | 0.00% | -0.11% | 0.02% | 0.00% | 97.579% | 100% |
| A3 | -0.03% | 0.02% | 0.02% | -0.03% | 97.019% | 100% |
| A4 | 0.03% | 0.10% | -0.04% | 0.03% | 97.039% | 100% |
| A5 | -0.04% | 0.20% | -0.09% | -0.03% | 97.165% | 99% |
+----------+--------+--------+--------+--------+------------+----------+

Add a new speed feature to analyze the stationarity of signal to prune
partition type searches.

The speed feature shows a good tradeoff of quality and speed up.

On speed 1, RA:
                Enc_Speedup  PSNR-YUV  Ratio
A1 (17frames)    +4.75%      +0.13%    36.5
A2 (33 frames)   +7.39%      +0.13%    56.8

STATS_CHANGED

- Consolidated initializations and moved early gating
  to the function start
- Introduced local variables for readability
- Renamed calc_masked_type_cost() to calc_comp_type_cost()
  and abstracted compound type cost calculation
- Replaced update_mask_best_mv() with av2_copy_array()
- Renamed rs2 to comp_type_rate
- Simplified the masked_compound_type_rd() signature
- Made cosmetic changes to follow coding guidelines

No Stats Changed

- Added CompTypeRdStats to consolidate RD cost members
  in COMP_RD_STATS
- Replaced local arrays in av2_compound_type_rd() with
  CompTypeRdStats

This improves readability, simplifies the function signature
and data copy in various helper functions.

No Stats Changed

- Unified pick_wedge() and pick_wedge_fixed_sign() to
  avoid code duplication
- Cleaned up the dead code and debug prints in pick_wedge()
- Modified get_wedge_cost() to cleanup unused parameter

No Stats Changed

- Modified comments to ensure correctness and correct punctuations
- Replaced INLINE with AVM_INLINE
- Replaced hard coded values with enums
- Used pre-increments in loop counters

No Stats Changed

Add default extra and encoder configuration options for
AVM_USAGE_REALTIME in av2/av2_cx_iface.c to match standard RTC encoding
defaults, and simplify test/rtc_test.cc to use the new defaults.

Setting `--rt` in the `avmenc` command line applies standard Realtime
(RTC) defaults equivalent to:
  
--bit-depth=8 --passes=1 --lag-in-frames=0 --auto-alt-ref=0
--enable-tpl-model=0 \
--enable-keyframe-filtering=0 --enable-deblocking=1
--enable-restoration=0 \
--enable-intra-edge-filter=0 --enable-flip-idtx=1 --enable-masked-comp=0
\
--enable-onesided-comp=0 --enable-interintra-comp=0
--enable-smooth-interintra=0 \
--enable-diff-wtd-comp=0 --enable-interinter-wedge=0
--enable-interintra-wedge=0 \
--enable-global-motion=0 --enable-warped-motion=0
--enable-smooth-intra=0 \
--enable-paeth-intra=0 --enable-cfl-intra=0 --force-video-mode=1
--enable-overlay=0 \
--enable-angle-delta=0 --enable-trellis-quant=0 --enable-qm=0
--use-intra-dct-only=1 \
--coeff-cost-upd-freq=2 --mode-cost-upd-freq=2 --mv-cost-upd-freq=3
--frame-parallel=0 \
--aq-mode=0 --deltaq-mode=0 --frame-boost=0 --noise-sensitivity=0
--cdf-update-mode=1 \
--reduced-reference-set=1 --enable-ref-frame-mvs=0 --erp-pruning-level=0
\
--use-ml-erp-pruning=0 --enable-ext-partitions=0 --enable-mrls=0
--enable-pc-wiener=0 \
--enable-wiener-nonsep=0 --enable-tip=0 --enable-bawp=0 --enable-cwp=0 \
--enable-imp-msk-bld=0 --enable-ist=0 --enable-inter-ist=0
--enable-inter-ddt=0 \
--enable-cctx=0 --enable-ibp=0 --max-drl-refmvs=0 --max-drl-refbvs=0
--enable-refmvbank=0 \
--enable-opfl-refine=0 --enable-ccso=0 --enable-lf-sub-pu=0
--enable-adaptive-mvd=0 \
--enable-flex-mvres=0 --enable-joint-mvd=0 --enable-refinemv=0
--enable-mvd-sign-derive=0 \
--enable-parity-hiding=0 --enable-warp-delta=0 --enable-warp-extend=0
--tile-columns=0 \
--gf-min-pyr-height=0 --gf-max-pyr-height=0 --enable-sdp=0
--min-gf-interval=8 \
--max-gf-interval=8 --explicit-ref-frame-map=1 --enable-bru=0
--enable-ext-seg=0 \
--enable-six-param-warp-delta=0 --enable-cdef=1
--enable-cdef-on-skip-txfm=0 \
--dpb-size=8 --use-inter-dct-only=0 --enable-fsc=1 --enable-idtx-intra=1
\
--enable-palette=1 --enable-intrabc-ext=0 --enable-intrabc=1
--cpu-used=6 \
    --enable-gdf=0 --max-reference-frames=3

Set tx_size_search_level = USE_FAST_RD, and add level 3 to
restrict_tx_partition_type_search to only test
TX_PARTITION_NONE and TX_PARTITION_SPLIT.

Also switch cdef_pick_method to CDEF_FAST_SEARCH_LVL3, until the
CDEF_PICK_FROM_Q model is further evaluated/improved.

Only affects realtime mode speed >= 6.

This brings back some quality with small speed loss: on average over rtc
set of vga and 720p clips, ~10% bdrate gain with ~4% encoder speed
slowdown. This is with the command line settings below with the qp range
{70 80 90 100 110 120 130 150 170}:

--passes=1 --bit-depth=8 --obu --lag-in-frames=0 --kf-min-dist=1000
--kf-max-dist=1000 --auto-alt-ref=0 --enable-tpl-model=0
--enable-keyframe-filtering=0 --enable-deblocking=1
--enable-restoration=0 --enable-intra-edge-filter=0 --enable-flip-idtx=1
--enable-masked-comp=0 --enable-onesided-comp=0
--enable-interintra-comp=0 --enable-smooth-interintra=0
--enable-diff-wtd-comp=0 --enable-interinter-wedge=0
--enable-interintra-wedge=0 --enable-global-motion=0
--enable-warped-motion=0 --enable-smooth-intra=0 --enable-paeth-intra=0
--enable-cfl-intra=0 --force-video-mode=1 --enable-overlay=0
--enable-angle-delta=0 --enable-trellis-quant=0 --enable-qm=0
--use-intra-dct-only=1 --coeff-cost-upd-freq=2 --mode-cost-upd-freq=2
--mv-cost-upd-freq=3 --frame-parallel=0 --aq-mode=0 --deltaq-mode=0
--frame-boost=0 --noise-sensitivity=0 --cdf-update-mode=1
--reduced-reference-set=1 --enable-ref-frame-mvs=0 --erp-pruning-level=0
--use-ml-erp-pruning=0 --enable-ext-partitions=0 --enable-mrls=0
--enable-pc-wiener=0 --enable-wiener-nonsep=0 --enable-tip=0
--enable-bawp=0 --enable-cwp=0 --enable-imp-msk-bld=0 --enable-ist=0
--enable-inter-ist=0 --enable-inter-ddt=0 --enable-cctx=0 --enable-ibp=0
--max-drl-refmvs=0 --max-drl-refbvs=0 --enable-refmvbank=0
--enable-opfl-refine=0 --enable-ccso=0 --enable-lf-sub-pu=0
--enable-adaptive-mvd=0 --enable-flex-mvres=0 --enable-joint-mvd=0
--enable-refinemv=0 --enable-mvd-sign-derive=0 --enable-parity-hiding=0
--enable-warp-delta=0 --enable-warp-extend=0 --tile-columns=0
--gf-min-pyr-height=0 --gf-max-pyr-height=0 --end-usage=q
--use-fixed-qp-offsets=0 --psnr --threads=1 --enable-sdp=0
--min-gf-interval=8 --max-gf-interval=8 --explicit-ref-frame-map=0
--enable-bru=0 --enable-ext-seg=0 --enable-six-param-warp-delta=0
--enable-cdef=1 --enable-cdef-on-skip-txfm=0 --dpb-size=8
--use-inter-dct-only=0 --enable-fsc=1 --enable-idtx-intra=1
--enable-palette=1 --enable-intrabc-ext=0 --enable-intrabc=0
--tune-content=0 --rt --cpu-used=6 --enable-gdf=0
--max-reference-frames=3 --max-q=qp ---min-q=qp

Promote tx stat pruning to Speed 3
The speed feature was at speed 4.

Enabling the speed feature at speed 3 shows
encoder speed up with acceptable coding loss.

New test on top of base commit
[5d628d8](AOMediaCodec@5d628d8)

Speed 3
Testset                  PSNR-YUV      EncSpeedUp     Ratio
A1 (17 frames)       0.14%              8.28%               59.1
A2 (33 frames)       0.17%              6.53%               38.4

Original test on top of base commit
AOMediaCodec@fe1bfde

Performance on RA speed 3
Testset           PSNR-YUV    EncSpeedUp
A1 (17 frames)    0.15%        6.55%
A2 (33 frames)    0.22%        5.68%

Besides, we tested the speed feature at speed 2, it does not
meet the bar for A1.

Performance on RA speed 2
Testset           PSNR-YUV    EncSpeedUp
A1 (17 frames)    0.16%        4.02%
A2 (33 frames)    0.13%        4.46%

STATS_CHANGED

By reducing the number of candidates, we achieve a considerable
speed up with accepted coding loss.

Result of reducing the count to 3 for speed 4:

TestSet          PSNR-YUV   EncTime    Ratio
A1 (17 frames)    0.10%      96.70%    34.1
A2 (33 frames)    0.07%      96.81%    47.0

Result of reducing the count to 3 for speed 3:

TestSet          PSNR-YUV   EncTime    Ratio
A1 (17 frames)    0.11%      96.87%    29.4
A2 (33 frames)    0.09%      96.41%    41.3

If we reduce the count to 2
The speedup/loss ratio is good for A2, but not good enough for A1.
On speed 4, RA, the performance is:

TestSet          PSNR-YUV   EncTime    Ratio
A1 (17 frames)    0.41%      92.76%    19.0
A2 (33 frames)    0.19%      93.44%    36.9

STATS_CHANGED

…iaCodec#5304)

The wet pass starts mode search from an unbounded rd limit, so its
pruning cannot bite until it finds a good candidate on its own. Start it
instead from the dry pass's rd for the same block and size, plus a
margin.

The two passes see different contexts, so the bound can be violated. A
block left with no valid mode is re-searched once with the original
limit.

Anchor: commit c77f169
Speed 4 (cpu-used=4): FG16 CTC (33 frames, class A1 and A2, RA)

```
  +------------+-------+-------+-------+-------+------+------+
  | Class      |     Y |    Cb |    Cr |  wAvg | Enc% | Dec% |
  +------------+-------+-------+-------+-------+------+------+
  | A1         | -0.30 | -0.29 | -0.38 | -0.30 |  100 |  100 |
  | A2         | -0.17 |  0.07 |  0.31 | -0.15 |  101 |  100 |
  | Avg w/o B2 | -0.21 | -0.04 |  0.11 | -0.19 |  100 |  100 |
  +------------+-------+-------+-------+-------+------+------+
```

STATS_CHANGED

- Introduce the reduce_max_drl_refmvs speed feature to reduce the
default DRL candidate count from 4 to 3 in higher encode speeds.
- Enable frame-level DRL adaptation (allow_frame_max_drl_bits = 1) when
reduce_max_drl_refmvs is active: key, boosted, and low pyramid level
frames (<= 2) retain 4 candidates (max_drl_bits = 3) to retain reference
quality, while leaf frames use 3 candidates (max_drl_bits = 2).
- Condition newmv_drl_search_limit on reduce_max_drl_refmvs to ensure
both ref_mv_idx 0 and 1 are searched before reusing results, preventing
motion predictor diversity collapse.
- The existing --max-drl-refmvs CLI flag can override this behavior.

Evaluation based on 2abcfff with 33 frames at speed 2:
```
+---------+--------+--------+--------+--------+----------+----------+
| Summary |   Y    |   U    |   V    |  YUV   | Enc-time | Dec-time |
+---------+--------+--------+--------+--------+----------+----------+
| A1      | +0.05% | -0.05% | +0.10% | +0.05% |  98.1%   | 100.0%   |
| A2      | +0.08% | +0.27% | -0.03% | +0.08% |  97.5%   | 100.5%   |
+---------+--------+--------+--------+--------+----------+----------+
```

Pulled common calculations from the pixel filtering loop and
conducted them beforehand.

CTC tests showed no stats change.

)

Added a speed feature to early terminate the warp
delta refinement loop if the rate-distortion cost
improvement in a pass drops below 5%.

Results for RA CTC, A1 17 frames, A2 33 frames,
speed 1: (Anchor: 7c9dfba)

```
+-----+-------+-------+-------+-------+---------------+
|Class|   Y   |   Cb  |   Cr  | wAvg  |EncInstCount(%)|
+-----+-------+-------+-------+-------+---------------+
| A2  |-0.0004|-0.0367|-0.2123|-0.0103| 98.93         |
| A1  |-0.0649| 0.1274| 0.0844|-0.0513| 98.89         |
+-----+-------+-------+-------+-------+---------------+
```
STATS_CHANGED for speed >= 1

Enable frame level bawp based on histogram. Enabled for speed 2 and
above.

33 frame RA results for speed 2 and 3 are as follows: 

+---------+--------+-------+-------+-------+----------+----------+
| Summary | Y | U | V | YUV | Enc-time | Dec-time |
+---------+--------+-------+-------+-------+----------+----------+
Speed2, Overall w/o B2 (RA) +0.00% +0.01% +0.01% +0.00% 99.51% 100.12%
Speed3, Overall w/o B2 (RA) +0.00% +0.01% +0.03% +0.00% 99.45% 99.53%

+---------+--------+-------+-------+-------+----------+----------+

…c#5314)

The expensive pixel-domain distortion path in search_tx_type()
  is skipped if the lightweight transform-domain RD estimate
  already exceeds best_rd or ref_best_rd. This is controlled by
  a new speed feature flag skip_pixel_dist_calc_using_tx_dist,
  enabled for speed >= 1.

  STATS_CHANGED for speed >= 1

 
  Test results (RA) for Speed 1
  Anchor: commit 5d628d8
  A1 - 17 frames
  A2 - 33 frames
    
  ```
  +------------+-------+-------+-------+-------+-------------+
  | Class      |     Y |    Cb |    Cr |  YUV  |EncInstCount%|
  +------------+-------+-------+-------+-------+-------------+
  | A1         | -0.03 |  0.43 | -0.06 |  0.00 |    95.84    |
  | A2         |  0.04 |  0.46 | -0.15 |  0.05 |    97.35    |
  +------------+-------+-------+-------+-------+-------------+
  ```