ECM-15.0 different intra results

Encoder produces different results for intra pictures.

Test 1. EncoderApp.exe -c encoder_intra_ecm.cfg -c ../cfg/per-sequence/RaceHorses.cfg --SEIDecodedPictureHash=1 -i RaceHorses.yuv -q 27 -fs 32 -f 33

Test 2. EncoderApp.exe -c encoder_intra_ecm.cfg -c ../cfg/per-sequence/RaceHorses.cfg --SEIDecodedPictureHash=1 -i RaceHorses.yuv -q 27 -fs 64 -f 1

POC4 of Test 1 and POC0 of Test 2 shall be the same, but they are different.

Test 1 log.

POC    0 LId:  0 TId: 0 ( IDR_N_LP, I-SLICE, QP 27 )      88392 bits [Y 39.5806 dB    U 39.6877 dB    V 40.7572 dB] [xY 4043ca515858e3a2 xU 4043d80531fd939e xV 404460ece5db75f0] [ET   170 ] [L0] [L1] [MD5:959aaeb90dcb15d78a4afa68f47c52df,3e03cb930928b26f6ea84f3773f0d671,16be35b0a28bfc03e6238f4f104a64cc]
POC    1 LId:  0 TId: 0 ( CRA, I-SLICE, QP 27 )      83528 bits [Y 39.6107 dB    U 40.1086 dB    V 41.2280 dB] [xY 4043ce2b7b927098 xU 40440de6c3c9a88b xV 40449d2e4a074d7b] [ET   156 ] [L0] [L1] [MD5:7a946301f3eafb571ca4d483c04c4519,8578a3265ed488d1f9b3345f1dd652d3,fdeb0b33a38dc1a37e00d2eaf827fc13]
POC    2 LId:  0 TId: 0 ( CRA, I-SLICE, QP 27 )      88584 bits [Y 39.6703 dB    U 39.9237 dB    V 40.9604 dB] [xY 4043d5cbc14a2904 xU 4043f63bd66179d7 xV 40447aeefa6b13de] [ET   172 ] [L0] [L1] [MD5:9e17a0c75a6483f54b9cba341cac92ef,89de50ae3a6f4def6534eeb00deac2c2,d3b5f833edf4aa62cb2b231e3ee40141]
POC    3 LId:  0 TId: 0 ( CRA, I-SLICE, QP 27 )      82232 bits [Y 40.0699 dB    U 40.3130 dB    V 40.9826 dB] [xY 404408f1f69b298a xU 4044281180f0d269 xV 40447dc71311bb8a] [ET   149 ] [L0] [L1] [MD5:e16ddbbce0301fcb1c1650131cdc5ed2,41d80b075dafd74c680974a6bd748b8a,4df420b90da79ce310c924384e9558bb]
POC    4 LId:  0 TId: 0 ( CRA, I-SLICE, QP 27 )      84960 bits [Y 39.8790 dB    U 40.2030 dB    V 40.9792 dB] [xY 4043f0836aa677c4 xU 404419fc87dd44b4 xV 40447d5530b089ac] [ET   166 ] [L0] [L1] [MD5:d0aeb7f29023b3746b81ec78f93d18df,6b8f31f088acd52b06a324a54bb7f1ca,087ae4d007dcde60cf3d3f938201af5f]

Test 2 log.

POC    0 LId:  0 TId: 0 ( IDR_N_LP, I-SLICE, QP 27 )      84752 bits [Y 39.8859 dB    U 40.1307 dB    V 41.1080 dB] [xY 4043f1669c25fdfa xU 404410bb318eb2c1 xV 40448dd17ed48117] [ET   169 ] [L0] [L1] [MD5:673c4946af61436b9802a0780c526d24,c0f3f74a2bd7ee4bb6813f0200b7d403,e444945251d3ef7312503b4807569cef]

When JVET_AJ0146_TIMDSAD from MR !773 (merged) is disabled then there is no issue in this test.

If to take MR !802 with added TIMD traces then the first difference is in TIMD SAD encoder decision.

Test 1 POC4.

cu_timd_flag() ctx=1 pos=(0,16) timd=0
eip_flag() pos=(0,16) mode=0
cu_tmrl_flag() ctx=0 pos=(0,16) tmrl=0
intra_subpartitions() etype=0 pos=(0,16) ispIdx=0
intra_luma_pred_modes() idx=0 pos=(0,16) pl_idx=0
intra_luma_pred_modes() idx=0 pos=(0,16) predIdx=0 mpm=1 secondmpm=0

Test 2 POC0.

cu_timd_flag() ctx=1 pos=(0,16) timd=1
cu_timd_flag() ctx=1 pos=(0,16) timdSad=1

@dbsansli, can you please take a look into this issue?