GitLab

remove SVN detection

See merge request jvet/VVCSoftware_BMS!113

With new Xcode and clang-10.0 SVN revision number detection in CMakeLists.txt causes a compilation error with cyclic redundancies.
- Fix by removing this feature completely. (useless in git repo)
- Not affecting results

Ternary split: fix for issue #87

See merge request jvet/VVCSoftware_BMS!111

Uses option #1 re-defining the meaning of maxTtSize being the longest side of the partition.

Cpr para seq fix

See merge request jvet/VVCSoftware_BMS!108

Revert "Merge branch 'High_Precision_Motion_Vector_Storage' into 'master'"

See merge request jvet/VVCSoftware_BMS!110

Revert "Merge branch 'cfgUpdate2' into 'master'"

See merge request jvet/VVCSoftware_BMS!109

This reverts merge request !106

This reverts merge request !107

Update cfg files

See merge request jvet/VVCSoftware_BMS!107

High precision motion vector storage

See merge request jvet/VVCSoftware_BMS!106

Cpr para fix

See merge request jvet/VVCSoftware_BMS!105

Fix a bug that lead to different result when the macro is on and off

Motion vecotrs are always stored with 1/16-pel precision; highprec flag,
setLow() and setHigh() functions are removed from the MV class; SPS
HighPrecision flag is also removed.

The following strategies are followed:

1. Motion vector data in the buffer are always stored in high precision.
2. Motion vector used for motion compensation are always in high precision.
3. Motion vector difference data are always in low precision.
4. Encoder side motion estimation always use low precision.
5. Motion vector bits calculation are always performed in low precision.
6. Candidates of motion vector prediction list are always in low precision.

Motion vecotrs are always stored with 1/16-pel precision and the High
precision motion vector SPS flags are removed.

The following strategies are followed:

1. Motion vector data in the buffer are always stored in high precision.
2. Motion vector used for motion compensation are always in high
precision.
3. Motion vector difference data are always in low precision.
4. Encoder side motion estimation always use low precision.
5. Motion vector bits calculation are always performed in low precision.
6. Candidates of motion vector prediction list are always in low
precision.

fix performance differences when CPR macro off and CPR on but config flag off

See merge request jvet/VVCSoftware_BMS!103

patch from bug report #84: encoder bug for affine 6-parameter search

See merge request jvet/VVCSoftware_BMS!102

K0248 GBI:  fix interaction with Affine mode

See merge request jvet/VVCSoftware_BMS!101

Inter AMT encoder fix

See merge request jvet/VVCSoftware_BMS!100

flag off

Patch provided by: Yuwen He

move tools stats after STATS__TOOL_TOTAL_FRAME, otherwise it is not compiling with JEM_TOOLS = 0 due to 0 array size

# Conflicts:
#	source/Lib/EncoderLib/EncCu.cpp
#	source/Lib/EncoderLib/EncCu.h

Motion vecotrs are always stored with 1/16-pel precision and the
following strategies are followed:

1. Motion vector data in the buffer are always stored in high precision.
2. Motion vector used for motion compensation are always in high
precision.
3. Motion vector difference data are always in low precision.
4. Encoder side motion estimation always use low precision.
5. Motion vector bits calculation are always performed in low precision.
6. Candidates of motion vector prediction list are always in low
precision.

use correct vector resolution in CPR search range setting

See merge request jvet/VVCSoftware_BMS!97

Remove bug fix macro under JVET_K0346

See merge request jvet/VVCSoftware_BMS!96

create BMS_TOOLS macro to switch all BMS only tools (including JEM_TOOLS)

See merge request jvet/VVCSoftware_BMS!95