Commit graph

4576 commits

Author SHA1 Message Date
Joose Sainio bbe96a3217 ???? ??? ???? 2023-04-28 15:55:55 +03:00
Joose Sainio ab4f63425d [avx2] This has worked but I'm pretty sure these should be unaligned 2023-04-26 14:41:04 +03:00
Joose Sainio bdd7851e86 [avx2] Simplify 2023-04-26 10:39:23 +03:00
Joose Sainio 473b1b8bb1 [avx2] AVX2 version of depquant now exactly matches scalar version 2023-04-24 08:10:35 +03:00
Joose Sainio 49a36dd455 [dep_quant] Fix rate_estimator and quant_block init cases 2023-04-21 11:42:42 +03:00
Joose Sainio d834a448c3 [avx2] Fix 2023-04-20 13:53:11 +03:00
Joose Sainio 65a875c733 [avx2] fix 2023-04-20 11:28:38 +03:00
Joose Sainio 35d0f58d51 WIP 2023-04-19 15:40:53 +03:00
Joose Sainio 30f1e18bc4 [avx2] replace or 2023-04-19 14:00:36 +03:00
Joose Sainio 8edcdd692b [avx2] Replace loads and stores with non-avx512 stores 2023-04-19 13:38:27 +03:00
Joose Sainio d0024169c2 [avx2] Replace _mm_and_epi32 with _mm_and_si128 2023-04-19 12:34:43 +03:00
Joose Sainio 0e5fb90278 [dep_quant] Clean up 2023-04-19 12:25:54 +03:00
Joose Sainio 9ab59fcc24 [depquant] update_state_eos_avx2 working 2023-04-18 15:43:30 +03:00
Joose Sainio 1ed989ffca WIP 2023-04-17 15:14:35 +03:00
Joose Sainio 61536cb40e [depquant] Only initialize rate_estimator when necessary 2023-04-17 14:45:55 +03:00
Joose Sainio e746db22ed [depquant] Initialize quant_block only when necessary 2023-04-17 14:18:57 +03:00
Joose Sainio b8878f029f [avx2] WIP update_state_eos_avx2 2023-04-17 13:52:58 +03:00
Joose Sainio 963db5a407 [avx2] and last 2023-04-14 09:55:09 +03:00
Joose Sainio 12fea6f8b1 [depquant] remove an unnecessary memcpy 2023-04-14 08:25:33 +03:00
Joose Sainio 6e2eaf9d6b [avx2] Do decision cost comparison with avx2 2023-04-13 15:20:49 +03:00
Joose Sainio fcbd12fef3 WIP 2023-04-12 15:36:45 +03:00
Joose Sainio c2ca1da666 [depquant] Pre calculate things
sig_ctx_offset gtx_ctx_offset cg_pos pos_y  pos_x next_sbb_right next_sbb_below
2023-04-12 15:02:06 +03:00
Joose Sainio d71eb77d1a [avx2] update_states_avx2 working 2023-04-12 10:41:37 +03:00
Joose Sainio 93f1726cba [avx2] WIP update_states_avx2 2023-04-10 15:31:05 +03:00
Joose Sainio 8d2a14f562 [avx2] simplify 2023-04-10 08:40:35 +03:00
Joose Sainio a57551c060 [avx2] check_rd_costs_avx2 done 2023-04-09 14:05:50 +03:00
Joose Sainio ba1b15e476 [avx2] WIP check_rd_costs_avx2, almost? 2023-04-08 18:58:40 +03:00
Joose Sainio a74332aa93 [avx2] WIP check_rd_costs_avx2 2023-04-07 15:15:58 +03:00
Joose Sainio 1f12e71a74 [dep_quant.c] Small refactor 2023-04-07 08:14:06 +03:00
Joose Sainio fc13b395c0 [depquant] AoS -> SoA pre quant 2023-04-06 16:07:38 +03:00
Joose Sainio 87c9d94484 [depquant] AoS -> SoA all states 2023-04-06 15:12:29 +03:00
Joose Sainio d178c72842 [depquant] AoS -> SoA for Decision 2023-04-05 11:17:16 +03:00
Joose Sainio 9120a04e77 [fix] Minor fixes 2023-04-05 09:33:00 +03:00
Joose Sainio 6f41929e5a [mtt] Fix small issues with luma and chroma searches 2023-03-31 13:37:22 +03:00
Joose Sainio 7022e44b5c [dual-tree] Fix chroma tree split model context derivation during search 2023-02-23 13:41:59 +02:00
Joose Sainio 49ca61c201 [dual-tree] Small fixes 2023-02-23 12:50:30 +02:00
Joose Sainio ca7cc7881f [dual-tree] Remove the limitation of not allowing 2 height chroma blocks in dual tree 2023-02-23 08:48:08 +02:00
Joose Sainio 899a0323fb [dual-tree] WIP simplification 2023-02-22 14:52:29 +02:00
Joose Sainio 3ce7691f6e [mtt] Fix various small issues and DepQuant for non-square blocks 2023-02-15 14:23:55 +02:00
Joose Sainio 1c84b1edeb [mtt] Minor fixes 2023-02-14 10:36:01 +02:00
Joose Sainio e54a451289 [mtt] Fix couple of issues with 64x32 CUs and non square tr skip rdoq 2023-02-08 14:50:03 +02:00
Joose Sainio b7905c7238 [ISP] Fix ISP cost calculation and DepQuant with mts 2023-02-03 14:58:36 +02:00
Joose Sainio fcb92849e0 [DepQuant] Fix isp+depquant and trskip + isp 2023-01-31 14:27:35 +02:00
Joose Sainio 44e4f72fb0 [DepQuant] Fix for mts and lfnst being quantized incorrectly during search 2023-01-23 13:39:22 +02:00
Joose Sainio e6b4cb5bb3 Merge branch 'depquant' of gitlab.tuni.fi:cs/ultravideo/vvc/uvg266 into depquant 2023-01-20 08:36:32 +02:00
Joose Sainio fd43064cfa [DepQuant] Isp and chroma 2023-01-19 16:30:47 +02:00
Joose Sainio 9f3df67718 [DepQuant] Fix 2023-01-17 15:18:13 +02:00
Joose Sainio 558b16baf1 [DepQuant] Fix 2023-01-17 11:03:14 +02:00
Joose Sainio dfda6810c5 [DepQuant] Working but not necessarily improving 2023-01-12 15:15:52 +02:00
Joose Sainio 49f6e72c00 [DepQuant] WIP: doesn't crash but bitstream is illegal and quality a lot worse 2023-01-11 14:17:18 +02:00