ECM merge requestshttps://vcgit.hhi.fraunhofer.de/ecm/ECM/-/merge_requests2024-03-28T13:53:22Zhttps://vcgit.hhi.fraunhofer.de/ecm/ECM/-/merge_requests/640fix JVET_AG0061_INTER_LFNST_NSPT: checking number of non-null coefficients wh...2024-03-28T13:53:22ZFabrice URBANfix JVET_AG0061_INTER_LFNST_NSPT: checking number of non-null coefficients when NumSignPred=0fix encoder crash when checking number of non-null coefficients when NumSignPred=0 for JVET_AG0061_INTER_LFNST_NSPT.
Number of non-null coefficients is also checked for chroma with "cctx(tu, **COMPONENT_Y**, signHiding)", and also for no...fix encoder crash when checking number of non-null coefficients when NumSignPred=0 for JVET_AG0061_INTER_LFNST_NSPT.
Number of non-null coefficients is also checked for chroma with "cctx(tu, **COMPONENT_Y**, signHiding)", and also for non-lfnst blocks.https://vcgit.hhi.fraunhofer.de/ecm/ECM/-/merge_requests/639Fix: Remove extra lines for LicDelta context2024-03-27T16:58:55ZVadim SereginFix: Remove extra lines for LicDelta contextremove extra lines for LicDeltaremove extra lines for LicDeltahttps://vcgit.hhi.fraunhofer.de/ecm/ECM/-/merge_requests/638fix JVET_AG0098 uninitialized value2024-03-25T10:36:26ZFabrice URBANfix JVET_AG0098 uninitialized valuefix Conditional jump or move depends on uninitialised value(s) in InterSearch::predInterSearch
and other encoder/decoder traces having unreliable valuesfix Conditional jump or move depends on uninitialised value(s) in InterSearch::predInterSearch
and other encoder/decoder traces having unreliable valueshttps://vcgit.hhi.fraunhofer.de/ecm/ECM/-/merge_requests/637Fix: JVET-AG0276 fixes2024-03-26T18:37:04ZFabrice URBANFix: JVET-AG0276 fixesfixed deactivation of encoder options affAltLM and MergeAffOppositeLic if affine is OFF (causing a decoder crash)
some code cleaning (macros OFF compilation and traces)fixed deactivation of encoder options affAltLM and MergeAffOppositeLic if affine is OFF (causing a decoder crash)
some code cleaning (macros OFF compilation and traces)https://vcgit.hhi.fraunhofer.de/ecm/ECM/-/merge_requests/636fix SPS::m_itmpLicMode not being initialized at decoder if IBC is OFF2024-03-22T09:58:12ZFabrice URBANfix SPS::m_itmpLicMode not being initialized at decoder if IBC is OFFfix SPS::m_itmpLicMode not being initialized at decoder if IBC is OFF
uninitialized variable reported by Valgrind. also creates decoder mismatch when compiling under "debug" configuration with visualfix SPS::m_itmpLicMode not being initialized at decoder if IBC is OFF
uninitialized variable reported by Valgrind. also creates decoder mismatch when compiling under "debug" configuration with visualhttps://vcgit.hhi.fraunhofer.de/ecm/ECM/-/merge_requests/635Optimization: reduce amount of array allocations2024-03-18T10:00:41ZFabrice Le Léannecfabrice.leleannec@interdigital.comOptimization: reduce amount of array allocationsOptimization: reduce amount of array allocationsOptimization: reduce amount of array allocationshttps://vcgit.hhi.fraunhofer.de/ecm/ECM/-/merge_requests/634fix missing xCalcDebCost2024-03-18T09:23:19ZTangi Poirierfix missing xCalcDebCostFix missing xCalcDebCost in functions:
xCheckRDCostIBCMode
xCheckRDCostInterIMV
xCheckRDCostInterMultiHyp2Nx2N
xCheckRDCostInterWoOBMC
Results are -0.01% in RA for 1s encodingFix missing xCalcDebCost in functions:
xCheckRDCostIBCMode
xCheckRDCostInterIMV
xCheckRDCostInterMultiHyp2Nx2N
xCheckRDCostInterWoOBMC
Results are -0.01% in RA for 1s encodinghttps://vcgit.hhi.fraunhofer.de/ecm/ECM/-/merge_requests/633Fix: activate ItmpLicExtension for screen content coding2024-03-13T17:44:08ZFabrice Le Léannecfabrice.leleannec@interdigital.comFix: activate ItmpLicExtension for screen content codingFix: activate ItmpLicExtension for screen content codingFix: activate ItmpLicExtension for screen content codinghttps://vcgit.hhi.fraunhofer.de/ecm/ECM/-/merge_requests/632Fix: RPR related fix for JVET-AG00982024-03-12T23:59:50ZKenneth AnderssonFix: RPR related fix for JVET-AG0098The collocated position can become outside the collocated picture when RPR is enabled. The fix is to scale the collocated position so its within the collocated picture.The collocated position can become outside the collocated picture when RPR is enabled. The fix is to scale the collocated position so its within the collocated picture.https://vcgit.hhi.fraunhofer.de/ecm/ECM/-/merge_requests/631Fix #63 correct missing default parameters for one cabac ctx introduces in ag...2024-03-27T17:06:37ZFranck GalpinFix #63 correct missing default parameters for one cabac ctx introduces in ag0276correct issue #63correct issue #63https://vcgit.hhi.fraunhofer.de/ecm/ECM/-/merge_requests/630Fix: ECM-12.0 decoder crash for --EnableTMTools=0 (#61)2024-03-07T16:57:19ZYan ZhangFix: ECM-12.0 decoder crash for --EnableTMTools=0 (#61)Fix issue #61 that when setting EnableTMTools to 0 decoder crashes.Fix issue #61 that when setting EnableTMTools to 0 decoder crashes.https://vcgit.hhi.fraunhofer.de/ecm/ECM/-/merge_requests/629Fix: Related to RPR for JVET-AG00982024-03-06T20:07:10ZKenneth AnderssonFix: Related to RPR for JVET-AG0098Hopefully final fix for RPR also encoder crash for LDB. The issue was located to JVET-AG0098 that used a collocated picture which was scaled. The fix is to use unscaledPic.Hopefully final fix for RPR also encoder crash for LDB. The issue was located to JVET-AG0098 that used a collocated picture which was scaled. The fix is to use unscaledPic.https://vcgit.hhi.fraunhofer.de/ecm/ECM/-/merge_requests/628Fix: Related to RPR for JVET-AG02762024-03-06T18:33:27ZKenneth AnderssonFix: Related to RPR for JVET-AG0276https://vcgit.hhi.fraunhofer.de/ecm/ECM/-/merge_requests/627Fix #62 update cabac dumper and cabactraining to handle ecm12 new model (rate...2024-03-06T18:37:47ZFranck GalpinFix #62 update cabac dumper and cabactraining to handle ecm12 new model (rateoffset per slice from AG0196)Cabac dumper and training are updated to handle the new parameters introduce by AG0196. Note that file format is changed and not compatible with the previous cabactraining.Cabac dumper and training are updated to handle the new parameters introduce by AG0196. Note that file format is changed and not compatible with the previous cabactraining.https://vcgit.hhi.fraunhofer.de/ecm/ECM/-/merge_requests/626Fix: RPR related for JVET-AG01642024-03-07T19:02:28ZKenneth AnderssonFix: RPR related for JVET-AG0164RPR has encoder decoder mismatch. The issue was located to JVET-AG0164. The reason was located to the check of scaled reference for GPM which before JVET-AG0164 always return false, e.g. always allow scaled reference. The fix is to align...RPR has encoder decoder mismatch. The issue was located to JVET-AG0164. The reason was located to the check of scaled reference for GPM which before JVET-AG0164 always return false, e.g. always allow scaled reference. The fix is to align that also with JVET-AG0164 so that scaled reference is allowed. After these fixes for AG0058, AG0145 and this fix there is hopefully always encoder decoder match also for RPR..https://vcgit.hhi.fraunhofer.de/ecm/ECM/-/merge_requests/625Fix: RPR related for JVET-AG01452024-03-06T18:40:13ZKenneth AnderssonFix: RPR related for JVET-AG0145RPR gives encoder decoder mismatch. One issue was located to JVET-AG0145. A fix is provided for the encoder so that the unscaled collocated picture is used to get the actual max and min values for that picture. Otherwise the delta parame...RPR gives encoder decoder mismatch. One issue was located to JVET-AG0145. A fix is provided for the encoder so that the unscaled collocated picture is used to get the actual max and min values for that picture. Otherwise the delta parameters are coded compared to 0 and 1023 (non correct max and min values) on encoder side but derived correctly on decoder side.https://vcgit.hhi.fraunhofer.de/ecm/ECM/-/merge_requests/624Fix: RPR related for JVET-AG00582024-03-07T16:53:07ZKenneth AnderssonFix: RPR related for JVET-AG0058Decoder stops when RPR functionality is tested. Issue was tracked down to JVET-AG0058. The provided fix omits use of scaled collocated picture for JVET-AG0058. Still encoder decoder mismatch. More MRs to come soon.Decoder stops when RPR functionality is tested. Issue was tracked down to JVET-AG0058. The provided fix omits use of scaled collocated picture for JVET-AG0058. Still encoder decoder mismatch. More MRs to come soon.https://vcgit.hhi.fraunhofer.de/ecm/ECM/-/merge_requests/623Fix: Add missing parameters copy in copySliceInfo for JVET_AG01452024-02-21T17:32:13ZKai Cuikai.cui@outlook.deFix: Add missing parameters copy in copySliceInfo for JVET_AG0145https://vcgit.hhi.fraunhofer.de/ecm/ECM/-/merge_requests/622Fix: correct regionId in bvRegionIdList.2024-02-20T18:51:26ZMilos RadosavljevicFix: correct regionId in bvRegionIdList.Fix regionId in bvRegionIdList. It is more correct to use regionId instead of checkIdx. Change does not affect the code execution and it should be bit-exact.Fix regionId in bvRegionIdList. It is more correct to use regionId instead of checkIdx. Change does not affect the code execution and it should be bit-exact.https://vcgit.hhi.fraunhofer.de/ecm/ECM/-/merge_requests/621Fix for decoder debug for JVET-AG00982024-02-20T19:28:57ZKenneth AnderssonFix for decoder debug for JVET-AG0098RPR has encoder and decoder mismatch with the master. To enable debugging of the RPR issue another issue needs to be fixed first. The debug and the release mode of the decoder give different outputs. On the encoder side "setAmvpSbTmvpEna...RPR has encoder and decoder mismatch with the master. To enable debugging of the RPR issue another issue needs to be fixed first. The debug and the release mode of the decoder give different outputs. On the encoder side "setAmvpSbTmvpEnabledFlag" is set to false when "Slice->getPicHeader()->getEnableTMVPFlag()" is false. This do not happen on the decoder side. The fix set "setAmvpSbTmvpEnabledFlag" to false when "Slice->getPicHeader()->getEnableTMVPFlag()" is false. This change makes debug and release mode to give same output so that further debugging can be made of the encoder decoder mismatch for RPR.
Example to see that RPR has encoder decoder mismatch can be detected by enabling config parameters RPR and RPRFunctionalityTesting on BQSquare in random access configuration. Suspects that issue is somehow related to JVET-AG0164. At least when trying to disable the macro JVET_AG0164_AFFINE_GPM the encoder stops before completion.