The flag is parsed incorrectly which causes decoder crash.

if (sei.m_constantPatchSizeFlag) { sei_read_flag(pDecodedMessageOutputStream, val, "nnpfc_constant_patch_size_flag"); sei.m_constantPatchSizeFlag = val; } else

Correct way to parse is as follows:

sei_read_flag(pDecodedMessageOutputStream, val, "nnpfc_constant_patch_size_flag"); sei.m_constantPatchSizeFlag = val; if (sei.m_constantPatchSizeFlag) {

sei_read_uvlc(pDecodedMessageOutputStream, val, "nnpfc_patch_width_minus1"); sei.m_patchWidthMinus1 = val;

sei_read_uvlc(pDecodedMessageOutputStream, val, "nnpfc_patch_height_minus1"); sei.m_patchHeightMinus1 = val;

} else

VTM-20 encoding crash was reported when slice level multi-qp optimization is used. The issue could be reproduced with following command line.

/EncoderAppStatic -c ../cfg/encoder_lowdelay_P_vtm.cfg -i ../../../sequences/RaceHorses_832x480_30.y4m -v 6 -dph 1 -q 37 -f 9 --DeltaQpRD=3

The encoding log is

VVCSoftware: VTM Encoder Version 20.0 [Linux][GCC 12.2.0][64 bit] [SIMD=AVX2]

Warning: Y4M file info is different from input setting. Using the info from Y4M file

* WARNING: For conforming bitstreams a valid Level value must be set! * Input File : ../../../sequences/RaceHorses_832x480_30.y4m Bitstream File : str.bin Reconstruction File : rec.yuv Real Format : 832x480 30Hz Internal Format : 832x480 30Hz Sequence PSNR output : Linear average only Hexadecimal PSNR output : Disabled Sequence MSE output : Disabled Frame MSE output : Disabled MS-SSIM output : Disabled Cabac-zero-word-padding : Enabled Frame/Field : Frame based coding Frame index : 0 - 8 (9 frames) Profile : main_10 AllRapPicturesFlag : 0 CTU size / min CU size : 128 / 4 subpicture info present flag : Disabled subpicture ID present flag : Disabled Max TB size : 64 Motion search range : 64 Intra period : -1 Decoding refresh type : 0 DRAP period : 0 EDRAP period : 0 QP : 37 Max dQP signaling subdiv : 0 Cb QP Offset (dual tree) : 0 (0) Cr QP Offset (dual tree) : 0 (0) QP adaptation : 0 (range=0) GOP size : 8 Input bit depth : (Y:8, C:8) MSB-extended bit depth : (Y:8, C:8) Internal bit depth : (Y:10, C:10) Chroma QP offset list : Disabled extended_precision_processing_flag : Disabled TSRC_Rice_present_flag : Disabled reverse_last_sig_coeff_enabled_flag : Disabled transform_skip_rotation_enabled_flag : Disabled transform_skip_context_enabled_flag : Disabled high_precision_offsets_enabled_flag : Disabled rrc_rice_extension_flag : Disabled persistent_rice_adaptation_enabled_flag: Disabled cabac_bypass_alignment_enabled_flag : Disabled Cost function: : Lossy coding (default) RateControl : 0 WeightedPredMethod : 0 GDREnabled : 0 Max Num Merge Candidates : 6 Max Num Affine Merge Candidates : 5 Max Num Geo Merge Candidates : 5 Max Num IBC Merge Candidates : 6

TOOL CFG: IBD:1 HAD:1 RDQ:1 RDQTS:1 RDpenalty:0 LQP:0 SQP:3 ASR:0 MinSearchWindow:8 RestrictMESampling:0 FEN:1 ECU:0 FDM:1 ESD:0 TransformSkip:1 TransformSkipFast:1 TransformSkipLog2MaxSize:5 ChromaTS:1 BDPCM:0 Tiles: 1x1 Slices: 1 MCTS:0 SAO:1 ALF:1 CCALF:1 MaxNumALFAPS 8 AlfapsIDShift 0 ConstantJointCbCrSignFlagWPP:0 WPB:0 PME:2 WaveFrontSynchro:0 WaveFrontSubstreams:1 ScalingList:0 TMVPMode:1 DQ:1 SignBitHidingFlag:0 RecalQP:0 TOOL CFG: LFNST:0 MMVD:1 Affine:1 AffineType:1 AdaptBypassAffineMe:0 PROF:1 SbTMVP:1 DualITree:1 IMV:1 BIO:0 LMChroma:1 HorCollocatedChroma:1 VerCollocatedChroma:0 MTS:1(explicit intra) SBT:1 ISP:1 SMVD:0 CompositeLTReference:0 Bcw:0 BcwFast:0 LADF:0 CIIP:1 Geo:0 AllowDisFracMMVD:1 AffineAmvr:0 AffineAmvrEncOpt:0 AffineAmvp:1 DMVR:0 MmvdDisNum:6 JointCbCr:1 ACT:0 PLT:0 IBC:0 HashME:0 WrapAround:0 VirtualBoundariesEnabledFlag:0 VirtualBoundariesPresentInSPSFlag:1 vertical virtual boundaries:[ ] horizontal virtual boundaries:[ ] Reshape:1 (Signal:SDR Opt:0 CSoffset:1) MRL:1 MIP:0 EncDbOpt:1 FAST TOOL CFG: LCTUFast:1 FastMrg:1 MaxMergeRdCandNumTotal:5 MergeRdCandQuotaRegular:4 MergeRdCandQuotaRegularSmallBlk:4 MergeRdCandQuotaSubBlk:2 MergeRdCandQuotaCiip:1 MergeRdCandQuotaGpm:8 PBIntraFast:1 IMV4PelFast:1 MTSMaxCand: 3(intra) 4(inter) ISPFast:0 AMaxBT:1 E0023FastEnc:1 ContentBasedFastQtbt:0 UseNonLinearAlfLuma:1 UseNonLinearAlfChroma:1 MaxNumAlfAlternativesChroma:8 TTFastSkip:31 TTFastSkipThr:1.075 FastLocalDualTree:2 RPR:0 TemporalFilter:4/0 SEI CTI:0 BIM:0 SEI FGC:0 SEI processing Order:0

Input ChromaFormatIDC = 4:2:0

Output (internal) ChromaFormatIDC = 4:2:0

Non-environment-variable-controlled macros set as follows:

RExtDECODER_DEBUG_BIT_STATISTICS = 0

RExtHIGH_BIT_DEPTH_SUPPORT = 0

RExtHIGH_PRECISION_FORWARD_TRANSFORM = 0

started @ Wed Apr 26 09:44:58 2023

POC 0 LId: 0 TId: 0 ( IDR_N_LP, I-SLICE, QP 33 ) 150880 bits [Y 33.9538 dB U 36.6393 dB V 38.0217 dB] [ET 295 ] [L0] [L1] [MD5:7401e26ecd7d5c7e37ba437e2cdd0b25,ade216e995f5f6d6fc4e4df5358d517a,14284139ce2ddf75292f21fbecb2b14d] POC 1 LId: 0 TId: 0 ( TRAIL, P-SLICE, QP 44 ) 8168 bits [Y 30.6804 dB U 35.9144 dB V 37.3392 dB] [ET 80 ] [L0 0c] [L1] [MD5:42e6c4ab0612c22c3cc895a706afb8e7,fda9fcf32588525f94dd13a4b7b1aab0,137d350bf696a85e88a7051a7d0c9df0] POC 2 LId: 0 TId: 0 ( TRAIL, P-SLICE, QP 41 ) 9816 bits [Y 30.1372 dB U 35.5666 dB V 36.8531 dB] [ET 106 ] [L0 1c 0] [L1] [MD5:692ad0b11529789da60fcc1aa14390a7,26b70afeed3eae00dced1a7de50a294e,73feca41cfa0e9201db1247b36e3cf02] POC 3 LId: 0 TId: 0 ( TRAIL, P-SLICE, QP 44 ) 8736 bits [Y 29.5723 dB U 35.3917 dB V 36.5509 dB] [ET 97 ] [L0 2c 1 0] [L1] [MD5:8ab4bddfa0c1e9d2c2189783a2cb643f,adde8e3c6146457f41385033926a317c,ab78558a2f6b285a58796e93a0cddb6b]

ERROR: In function "intra_luma_pred_modes" in /usr/local/google/home/xlxiangli/program/g-vtm/source/Lib/EncoderLib/CABACWriter.cpp:1055: use of non-MPM

The ticket is about the text in the VVC spec in section 8.4.5.2.14 on pages 226 and 227, just after the equation 359. The text after eqn 359 in the spec should be updated to make it clear that luma samples prior to the in-loop filtering process are being used; and not luma samples prior to the deblocking filter.

Specifically, in the text from the spec, the "prior to the deblocking filter" should be changed to "prior to the in-loop filter process" at three places in the spec. On checking the software code, the samples prior to the in-loop filter process are being used. But the spec mentions prior to the deblocking filter. Suggestion is to update this.

As an aside, the text preceding eqn. 316 lists: The refW + refH + 1 + ( 2 * refIdx ) neighbouring samples refUnfilt[ x ][ y ] that are constructed samples prior to the inloop filter process, ... which seems the correct way to describe the text following eqn 359 (note: text near eqn. 316 seems correct, no issues in that).

Text following eqn. 359:

1. The collocated luma samples pY[ x ][ y ] with x = 0..nTbW * SubWidthC − 1, y= 0..nTbH * SubHeightC − 1

are set equal to the reconstructed luma samples prior to the deblocking filter process at the locations ( xTbY + x, yTbY + y ).

2. The neighbouring luma samples pY[ x ][ y ] are derived as follows:

– When availL is equal to TRUE, the neighbouring luma samples pY[ x ][ y ] with x = −3..−1, y = ( availT ? −1 : 0 )..SubHeightC * Max( nTbH, numSampL ) − 1, are set equal to the reconstructed luma samples prior to the deblocking filter process at the locations ( xTbY + x, yTbY +y ). – When availT is equal to FALSE, the neighbouring luma samples pY[ x ][ y ] with x = −2..SubWidthC * nTbW − 1, y = −2..−1, are set equal to the luma samples pY[ x ][ 0 ] – When availT is equal to TRUE, the neighbouring luma samples pY[ x ][ y ] with x = ( availL ? −1 : 0 )..SubWidthC * Max( nTbW, numSampT ) − 1, y = −3..−1, are set equal to the reconstructed luma samples prior to the deblocking filter process at the locations ( xTbY+ x, yTbY + y ). – When availL is equal to FALSE, the neighbouring luma samples pY[ x ][ y ] with x = −1, y = −2..SubHeightC * nTbH − 1, are set equal to the reconstructed luma samples pY[ 0 ][ y ].