DRAFT: Rebase av2enc on main - #5331

Open

yeqing-wu wants to merge 205 commits into

AOMediaCodec:av2-encfrom

yeqing-wu:rebase_av2enc_on_main

Commits on Jul 10, 2026

Commits on Jul 13, 2026

Clean up unused ref_frame_sign_bias[] in `AV2Common` structure (#5097)
satheeshkumar02
authored
160c6a7

Commits on Jul 14, 2026

Commits on Jul 15, 2026

Commits on Jul 16, 2026

Fix setting in multi_layers tests (#5117)
marpan-max
authored
7802259

Commits on Jul 18, 2026

Fix typo during checks for allowing extended sdp

ram-mohan
authored and
urvangjoshi
committed
229f4f9
Unify extended partition search functions

ram-mohan
authored and
urvangjoshi
committed
1682701
Refactor ext partition search in av2_rd_pick_partition

ram-mohan
authored and
urvangjoshi
committed
7821330
Refactor av2_rd_pick_partition()

ram-mohan
authored and
urvangjoshi
committed
431c17f
Fix partition timing data collection

ram-mohan
authored and
urvangjoshi
committed
636fcda

Commits on Jul 27, 2026

Check the requirement on is_first_tile_group (#5143)
wantehchang
authored
63be727

Commits on Aug 1, 2026

Assert MAX_PROFILES & PROFILE_BITS are consistent (#5182)
wantehchang
authored
2ed883d
Add the void cast to the tp_orig variable itself (#5174)

jzern
and
wantehchang
authored
48a7a18
extract_proto.cc: prefer std::size over ABSL_ARRAYSIZE (#5175)
jzern
authored
3efe9fa

Commits on Aug 3, 2026

Remove unneeded filtering of avmdec and avmenc

wantehchang
authored and
urvangjoshi
committed
94d3c8e
Declare two fuctions static

wantehchang
authored and
urvangjoshi
committed
ec3b746
Add NEON for highbd warp affine filter (#5169)
jjustiss-apple
authored
e838568
Remove unneeded av2_get_chroma_format_idc() calls

wantehchang
authored and
urvangjoshi
committed
1633b4a
av2_get_chroma_format_idc: more monochrome checks

wantehchang
authored and
urvangjoshi
committed
fa5e6f5

Commits on Aug 4, 2026

Commits on Aug 5, 2026

Commits on Aug 6, 2026

Refactor write_tile_info_max_tile (#5221)
wantehchang
authored
4b04e73
Simplify the second parameter of reuse_tile_params (#5223)
wantehchang
authored
12ab7c6
Add NEON for highbd inv_txfm (#5205)
jjustiss-apple
authored
2969f22

Commits on Aug 10, 2026

Commits on Aug 11, 2026

Commits on Aug 12, 2026

Partially undo avm_highbd_upsampled_pred_c changes (#5245)

wantehchang
and
urvangjoshi
authored
316dac3

Commits on Aug 13, 2026

avm_highbd_upsampled_pred_c: Set temp to comp_pred (#5259)
wantehchang
authored
c9431c4

Commits on Aug 17, 2026

decoder: check profile interop conformance ASAP (#5181)
wantehchang
authored
c3d517b

Commits on Aug 18, 2026

check_frame_unit_data returns 0 if vcl_found false (#5277)
wantehchang
authored
8a2a0a9

Commits on Aug 19, 2026

Remove local var bfirst in check_frame_unit_data() (#5278)
wantehchang
authored
4cdeb66

Commits on Aug 20, 2026

Commits on Aug 21, 2026

Fix encoder segfault when a GF group is truncated to one frame (#5302)
xinzhao-apple
authored
4bdc679

Commits on Aug 25, 2026

Use encoder-only method for reduced-tx-part-set option (#5313)
urvangjoshi
authored
cf38569
Add NEON for highbd convolve_2d_sr, convolve_x_sr, convolve_y_sr (#5312)
jjustiss-apple
authored
868edb0
Do not enable simple_motion_search_prune_rect (#5311)

xinzhao-apple
and
Xin Zhao
authored
493a008

Commits on Aug 26, 2026

Rework speed features 2 - 5 (#5080)

chengchen-google
authored and
yeqing-wu
committed
e7c2e35
Move set_sb_size() to encoder dir (#5084)

jingninghan
authored and
yeqing-wu
committed
00e6f49
Move get_partition() to enc dir (#5091)

jingninghan
authored and
yeqing-wu
committed
7712b85
Partitioning bugfix (#5095)

authored and
yeqing-wu
committed
a78da9e
Inter mode search: restructure loop, prune_motion_mode (#5098)
yeqing-wu
committed
6eb1009
Remove FSC path from intra joint UV function (#5111)

jingninghan
authored and
yeqing-wu
committed
9a98ca1
Rework single-ref to compound pruning (#5099)
yeqing-wu
committed
46ce624
Fix latent assert in handle_newmv MV-precision cache lookup (#5114)
yeqing-wu
committed
5056908
Add per-ref RD-based pruning for single-ref NEWMV / WARP_NEWMV (#5100)
yeqing-wu
committed
47f964b
Single-ref NEWMV DRL search speed features (#5101)
yeqing-wu
committed
ada5dd7
Add pruning for refinemv, INTERINTRA, and WARP_DELTA (#5102)
yeqing-wu
committed
7201c3c
Skip IST/STX cascade for intra-luma zero-quant primary (#5103)
yeqing-wu
committed
f46e69e
Rework av2_encode_block_intra_joint_uv (#5118)

jingninghan
authored and
yeqing-wu
committed
8056498
Remove mlayer/implicit-ref-map TIP suppression (#5119)
yeqing-wu
committed
3b750ab
Add cctx type array update back (#5123)

jingninghan
authored and
yeqing-wu
committed
ffa9264
Skip temporary prediction for opfl modes. (#5126)

jadhuran
authored and
yeqing-wu
committed
c666781
Hybrid search of multi way transform in rough search and winner mode search (#5125)

jadhuran
authored and
yeqing-wu
committed
38b6c95
Add a speed level for intra region in inter frames (#5130)

leolzhao
authored and
yeqing-wu
committed
a421252
Add 6 param warp search in winner mode by tid (#5129)

jadhuran
authored and
yeqing-wu
committed
8c38c92
Disable switchable SMVR at speed >= 1 and rename DMVR as SMVR or refinemv (#5134)

kslu-aom
authored and
yeqing-wu
committed
61dd959
Enhanced inter mode cache reuse (#5133)

leolzhao
authored and
yeqing-wu
committed
9e44a60
Refactor search_tx_type() (#5136)

mudassir-sg
authored and
yeqing-wu
committed
7180f49
Increase TCQ threshold based on base QP (#5132)

motongxu
authored and
yeqing-wu
committed
f7f6b5b
Add a speed feature to improve intra mode pruning

ranjit-tulabandu
authored and
yeqing-wu
committed
61f689a
Enable skip tx search when max_eob<512 (#5144)

motongxu
authored and
yeqing-wu
committed
31b7299
Encoder speedup for pruning partitions

linzhen-beep
authored and
yeqing-wu
committed
6ab2b41
Add realtime(rt) flag, encoder options / config

jianj-g
authored and
yeqing-wu
committed
dcffe9a
Speed up on uneven 4way partition and MV precision. (#5149)

jadhuran
authored and
yeqing-wu
committed
d7a266e
Add SSE2 optimization for interp_cubic() (#5135)

deepa-kg
authored and
yeqing-wu
committed
6b587a1
Transform partition encoder speedup (#5165)

linzhen-beep
authored and
yeqing-wu
committed
a8999fd
Pruning compound reference frame search in speed >= 1 (#5163)

kslu-aom
authored and
yeqing-wu
committed
d5d4752
Add macro for maximum qp selection (#5145)

arnab-dey-12
authored and
yeqing-wu
committed
802013f
Fast method for AMVD with newmv mode (#5166)

leolzhao
authored and
yeqing-wu
committed
b17b42a
Skip AMVD modes for NEAR NEWMV, NEW NEARMV modes (#5167)

jadhuran
authored and
yeqing-wu
committed
a4c1b1f
Add warp interintra in winner based on wrl index (#5164)

jadhuran
authored and
yeqing-wu
committed
888d484
Improve readability of search_tx_type() (#5160)

mudassir-sg
authored and
yeqing-wu
committed
01d2e36
MLP intra mode pruning (#5162)

YixinDu1
authored and
yeqing-wu
committed
1d12946
Cleanup PartitionSearchState context (#5157)

ram-mohan
authored and
yeqing-wu
committed
ff8d9d0
Use MFMV_STACK_SIZE as `process_ref` array size

satheeshkumar02
authored and
yeqing-wu
committed
a076a37
Refactor check_and_add_process_ref()

satheeshkumar02
authored and
yeqing-wu
committed
788833c
Remove unused `ProcessRefTMVP->type` and associated code

satheeshkumar02
authored and
yeqing-wu
committed
a01d4c7
Prevent auto inlining of motion_field_projection_side()

satheeshkumar02
authored and
yeqing-wu
committed
409e175
Adaptive tx search level (#5170)

motongxu
authored and
yeqing-wu
committed
de5f844
Rework av2_update_nbr_diagonal_c (#5172)

jingninghan
authored and
yeqing-wu
committed
3bb453d
MLP-based partition pre-screener (#5168)

YixinDu1
authored and
yeqing-wu
committed
6ada811
Add default speed features for RTC (#5155)

jianj-g
authored and
yeqing-wu
committed
b9be7ce
Simplify pc tree recurse functions

ram-mohan
authored and
yeqing-wu
committed
0a3f553
Tweak adaptive_tx_type_search speed feature

linzhen-beep
authored and
yeqing-wu
committed
865d74e
Add enable_low_complexity_decode (LC) mode (#5183)

yunqingwang1
authored and
yeqing-wu
committed
819fe8d
Cleanup unused code in compound type search (#5176)

RemyaPrakasan
authored and
yeqing-wu
committed
edee28c
Enable deblock filter for partition research (#5180)

jianle973
authored and
yeqing-wu
committed
2e7096f
Fix typo in collect_mv_stats_sb while accessing subtree

arnab-dey-12
authored and
yeqing-wu
committed
5ddfe21
Simplify ptree recurse functions

arnab-dey-12
authored and
yeqing-wu
committed
799fe93
Fix array subscript out of bounds warning in rdopt.c

jianj-g
authored and
yeqing-wu
committed
fd75e08
Extend perform_coeff_opt qindex threshold (#5196)

YixinDu1
authored and
yeqing-wu
committed
b5c720f
Extend prune_inter_tx_part_rd_eval qindex threshold (#5197)

YixinDu1
authored and
yeqing-wu
committed
802097c
Reduce winner mode transform candidates to 2 (#5198)

jadhuran
authored and
yeqing-wu
committed
9d14fb3
Improve AVX2 of av2_wedge_sse_from_residuals()

deepa-kg
authored and
yeqing-wu
committed
22629e6
Add AVX2 implementation for avm_highbd_comp_avg_upsampled_pred()

deepa-kg
authored and
yeqing-wu
committed
e7baf42
Optimize AVX2 of highbd SAD kernels

deepa-kg
authored and
yeqing-wu
committed
e106525
Avoid repeated function calls in av2_setup_motion_field()

satheeshkumar02
authored and
yeqing-wu
committed
d5bc0d4
Add helper function add_nearest_past_future_ref()

satheeshkumar02
authored and
yeqing-wu
committed
5702f42
Abstract TIP related code in av2_setup_motion_field()

satheeshkumar02
authored and
yeqing-wu
committed
617888c
Add helper function sort_and_update_ref_idx()

satheeshkumar02
authored and
yeqing-wu
committed
1e4a4ce
Remove unused segmentation functions (#5203)

arnab-dey-12
authored and
yeqing-wu
committed
7453ebc
Simplify tcq avx2 functions (#5208)

jingninghan
authored and
yeqing-wu
committed
571e413
Bugfix for deblocking filter in partition search: fix distortion scaling (#5212)

urvangjoshi
authored and
yeqing-wu
committed
32f6cbf
Speed up warped-motion refinement at speed 4 (#5206)

YixinDu1
authored and
yeqing-wu
committed
5c29568
Prune evaluation of compound modes using estimated RD-Cost (#5211)

ranjit-tulabandu
authored and
yeqing-wu
committed
4cda07f
Enable DC-block prediction at speed 4 and 5 (#5207)

YixinDu1
authored and
yeqing-wu
committed
4f5ab08
Reduce NEWMV DRL search limit at speed >= 4 (#5218)

authored and
yeqing-wu
committed
17cff57
Disable deblocking in partition search for speed >= 3 (#5219)

YixinDu1
authored and
yeqing-wu
committed
9f35bf5
Extend ref_mv_idx search pruning to speed 1-5 (#5220)

YixinDu1
authored and
yeqing-wu
committed
d83b451
Simplify tx_class related code in tcq avx2 (#5217)

jingninghan
authored and
yeqing-wu
committed
6adb582
Add tools-off options for low complexity decoding (#5213)

yunqingwang1
authored and
yeqing-wu
committed
ae87510
Optimize AV2 encoder by removing redundant memset calls (#5194)

JimDBh
authored and
yeqing-wu
committed
5fe2776
Revert "Optimize AV2 encoder by removing redundant memset calls (#5194)" (#5232)

urvangjoshi
authored and
yeqing-wu
committed
03bc9fe
Move lossless DPCM check inside prune_intra_y_mode

mudassir-sg
authored and
yeqing-wu
committed
493fd0f
Implement real-time variance-based partitioning

jianj-g
authored and
yeqing-wu
committed
3ec6d52
Prune uneven 4A/4B partitions with neighbor block boundaries (#5193)

urvangjoshi
authored and
yeqing-wu
committed
08b90b9
Code cleanup: remove partition arg from pick_sb_modes (#5229)

urvangjoshi
authored and
yeqing-wu
committed
35a42ed
Fix cost calculation in rectangular partition search

ram-mohan
authored and
yeqing-wu
committed
8690e18
Deduplicate boundary scans in extended-partition pruning

ram-mohan
authored and
yeqing-wu
committed
e54b66f
Refactor rd_try_subblock_new for reuse in partition search

ram-mohan
authored and
yeqing-wu
committed
e3c4178
Refactor search intra region partitioning function

ram-mohan
authored and
yeqing-wu
committed
1ca6a86
Refactor rectangular split partition search function

ram-mohan
authored and
yeqing-wu
committed
d497126
Refactor none_partition_search for consistency

ram-mohan
authored and
yeqing-wu
committed
ca39566
Refactor partition split search function

ram-mohan
authored and
yeqing-wu
committed
f71f83c
Apply fast method to new_newmv and new_newmv_opfl amvd sub-mode (#5238)

authored and
yeqing-wu
committed
f5532be
CCSO early terminate by cost threshold (#5237)

jadhuran
authored and
yeqing-wu
committed
b1d1de4
set_params_rd_pick_inter_mode: Remove dead code. (#5230)

urvangjoshi
authored and
yeqing-wu
committed
9b6e005
Rename `dist_wtd` and related terms to `cwp` (#5228)

satheeshkumar02
authored and
yeqing-wu
committed
624de3e
Add an aggressive low-complexity level

yunqingwang1
authored and
yeqing-wu
committed
40ae2fa
Add encoder options for aggressive low-complexity level

yunqingwang1
authored and
yeqing-wu
committed
29e4218
Increase exhaustive search threshold for speed 1

deepa-kg
authored and
yeqing-wu
committed
e2fe2c3
Implement non-RD partition evaluation for real-time mode (#5236)

jianj-g
authored and
yeqing-wu
committed
2be2cb3
Fix "Optimize AV2 encoder by removing redundant memset calls" (#5255)

JimDBh
authored and
yeqing-wu
committed
40e958d
Remove *Large from RTC test (#5257)

jianj-g
authored and
yeqing-wu
committed
6ee6a7a
Revert the unintended change under COLLECT_TX_SIZE_DATA (#5251)

mudassir-sg
authored and
yeqing-wu
committed
aa90cea
Increase exhaustive search threshold for speed >=2 (#5252)

deepa-kg
authored and
yeqing-wu
committed
b8e165a
Add reference-frame-selective warp search speed feature

JimDBh
authored and
yeqing-wu
committed
87bfe3a
Fix format

JimDBh
authored and
yeqing-wu
committed
c2409e8
Optimize speed features to disable H / 4-way Partitions (#5254)

urvangjoshi
authored and
yeqing-wu
committed
2950bcb
Use high_precision_mv_usage = LAST_MV_DATA for all speeds (#5266)

urvangjoshi
authored and
yeqing-wu
committed
ace3aac
mv_sf tuning for speed >= 1 (#5265)

linzhen-beep
authored and
yeqing-wu
committed
c9400b7
Disable extended SDP at speed 2 (#5260)

motongxu
authored and
yeqing-wu
committed
011f98e
Disable enable_winner_mode_for_coeff_opt (#5267)

jianle973
authored and
yeqing-wu
committed
107795f
Force max pb aspect ratio to be 4 at speed 4 (#5246)

motongxu
authored and
yeqing-wu
committed
c8f4bdb
Further increase exhaustive search threshold for speed 1 (#5263)

deepa-kg
authored and
yeqing-wu
committed
ec20308
Enable partition-deblock-rd for speed 3-6 (#5271)

jianle973
authored and
yeqing-wu
committed
7f656f6
Enable TCQ for all speed levels (#5270)

jadhuran
authored and
yeqing-wu
committed
d9a6a35
Add a speed feature to prune ref_mv_idx for WARP_NEWMV (#5269)

ranjit-tulabandu
authored and
yeqing-wu
committed
0da3756
Two-pass superblock partition search (#5253)
yeqing-wu
committed
45e5b21
Refactor ccso search (#5268)

ram-mohan
authored and
yeqing-wu
committed
171fec0
Fast warp delta search for four and six param models in warp newmv (#5272)

jadhuran
authored and
yeqing-wu
committed
e8c204a
Disable a speed feature to recover coding gains at sp3 (#5281)

chengchen-google
authored and
yeqing-wu
committed
e062332
Allow MV precision based on last frame MV data for < 4K resolution. (#5280)

urvangjoshi
authored and
yeqing-wu
committed
d20e3b3
Enable dc block pred level 1 at Speed 3 (#5283)

chengchen-google
authored and
yeqing-wu
committed
e1e616f
Disable a speed feature to recover coding gains at sp4 (#5285)

chengchen-google
authored and
yeqing-wu
committed
f37b2f7
Refactor handle_inter_mode and separate single and compound mode searches (#5239)

kslu-aom
authored and
yeqing-wu
committed
572c351
CCSO BO-first warm start under early termination (#5286)

chengchen-google
authored and
yeqing-wu
committed
7b3bb26
Bypass Luma plane search in CCSO (#5273)

authored and
yeqing-wu
committed
12e5b4a
Handle near zero coefficients in tcq (#5261)

jingninghan
authored and
yeqing-wu
committed
c7b93c8
Promote feature DC-block-pred-level to speed 3 (#5287)

chengchen-google
authored and
yeqing-wu
committed
8756061
Fix non-monotonic behavior for screen content (#5292)

yunqingwang1
authored and
yeqing-wu
committed
42ae3a0
Disable three speed-4 features that cost quality and save little time (#5288)

chengchen-google
authored and
yeqing-wu
committed
4abfd77
rtc: Reuse pc_tree in non-RD partition search (#5284)

jianj-g
authored and
yeqing-wu
committed
2a9136d
Consolidate intra FSC mode search into outer loop

SamuthirikaS
authored and
yeqing-wu
committed
b59e9df
Abstract intra mode pruning logic into a function

SamuthirikaS
authored and
yeqing-wu
committed
cc2c753
Abstract multi-winner mode processing into helper function

SamuthirikaS
authored and
yeqing-wu
committed
739a071
Rename variables for intra mode search

SamuthirikaS
authored and
yeqing-wu
committed
8cb7d44
Refactor av2_rd_pick_intra_sby_mode

SamuthirikaS
authored and
yeqing-wu
committed
2d97d8d
Abstract intra mode cost calculation into a function

SamuthirikaS
authored and
yeqing-wu
committed
6804210
Simplify av2_trellis_quant function (#5295)

jingninghan
authored and
yeqing-wu
committed
8e078b4
Promote WienerNS zero-refine to speed 1 (#5293)

chengchen-google
authored and
yeqing-wu
committed
ec7975d
Skip unit size search for Loop Restoration based on pyramid level for speed 4 (#5294)

tianqitliu
authored and
yeqing-wu
committed
5a4b089
Transform partition stationarity pruning (#5274)

chengchen-google
authored and
yeqing-wu
committed
b948286
Refactor av2_compound_type_rd

RemyaPrakasan
authored and
yeqing-wu
committed
c367721
Consolidate RD cost related members in COMP_RD_STATS

RemyaPrakasan
authored and
yeqing-wu
committed
750e584
Unify pick_wedge() and pick_wedge_fixed_sign()

RemyaPrakasan
authored and
yeqing-wu
committed
7418a50
Cosmetic changes in compound_type.c

RemyaPrakasan
authored and
yeqing-wu
committed
0d4122e
Set default configuration for realtime usage (--rt) (#5299)

jianj-g
authored and
yeqing-wu
committed
772a9d9
Adjust speed features for RTC speed >= 6. (#5300)

marpan-max
authored and
yeqing-wu
committed
70b9102
Promote tx stat pruning to Speed 3 (#5282)

chengchen-google
authored and
yeqing-wu
committed
b0d3d31
Reduce intra mode top count to 3 for speed 3 and 4 (#5289)

chengchen-google
authored and
yeqing-wu
committed
55d02ed
Bound the two-pass wet-pass mode search with the dry pass's rd (#5304)
yeqing-wu
committed
f41bc1f
Conditionally reduce max DRL ref MVs for speeds >= 2 (#5279)

kslu-aom
authored and
yeqing-wu
committed
a5ab19f
Temporal filter function optimization

yunqingwang1
authored and
yeqing-wu
committed
3c10235
Promote the uneven-4-way partition disable to speed 1 (#5305)

chengchen-google
authored and
yeqing-wu
committed
60cfe78
Add early termination in warp delta refinement (#5309)

deepa-kg
authored and
yeqing-wu
committed
c39b7f1
Add frame level on/off bawp (#5156)

jianle973
authored and
yeqing-wu
committed
63bf361
Skip pixel-domain distortion using tx-domain RD estimate (#5314)

ranjit-tulabandu
authored and
yeqing-wu
committed
b579dae

Uh oh!

Commits

Commits on Jul 6, 2026

Commits on Jul 7, 2026

Commits on Jul 8, 2026

Commits on Jul 9, 2026

Commits on Jul 10, 2026

Commits on Jul 13, 2026

Commits on Jul 14, 2026

Commits on Jul 15, 2026

Commits on Jul 16, 2026

Commits on Jul 18, 2026

Commits on Jul 27, 2026

Commits on Aug 1, 2026

Commits on Aug 3, 2026

Commits on Aug 4, 2026

Commits on Aug 5, 2026

Commits on Aug 6, 2026

Commits on Aug 10, 2026

Commits on Aug 11, 2026

Commits on Aug 12, 2026

Commits on Aug 13, 2026

Commits on Aug 17, 2026

Commits on Aug 18, 2026

Commits on Aug 19, 2026

Commits on Aug 20, 2026

Commits on Aug 21, 2026

Commits on Aug 25, 2026

Commits on Aug 26, 2026