Updated medium preset, vectorized DQ, improved SCC detection, fixes... (

#345)
fraunhoferhhi · Jan 18, 2024 · 8fe23c7 · 8fe23c7
1 parent c2ace2a
commit 8fe23c7
Show file tree

Hide file tree

Showing 51 changed files with 2,295 additions and 1,286 deletions.
diff --git a/cfg/experimental/lowdelay_medium.cfg b/cfg/experimental/lowdelay_medium.cfg
@@ -70,7 +70,7 @@ DualITree                     : 1      # separate partitioning of luma and chrom
 MinQTLumaISlice               : 8
 MinQTChromaISliceInChromaSamples : 4   # minimum QT size in chroma samples for chroma separate tree
 MinQTNonISlice                : 8
-MaxMTTDepth                   : 221111
+MaxMTTDepth                   : 1
 MaxMTTDepthI                  : 2
 MaxNumMergeCand               : 6      # Maximum number of merge candidates
 
@@ -130,7 +130,7 @@ ContentBasedFastQtbt          : 1      # Signal based QTBT speed-up: 0: disabled
 PBIntraFast                   : 1      # Intra mode pre-check dependent on best Inter mode, skip intra if it is not probable (0:off ... 2:fastest)
 FastQtBtEnc                   : 1      # Fast encoding setting for QTBT
 FastHAD                       : 0      # Use fast sub-sampled hadamard for square blocks >=32x32
-FastMrg                       : 3      # Fast methods for inter merge: 0: disabled, 1: vtm, 2: fast, 3: faster
+FastMrg                       : 2      # Fast methods for inter merge: 0: disabled, 1: vtm, 2: fast, 3: faster
 FastLocalDualTreeMode         : 1      # Fast intra pass coding for local dual-tree in intra coding region: 0: disable, 1: use threshold, 2: one intra mode only
 FastSubPel                    : 1      # Fast sub-pel ME: 0: disabled, 1: enabled
 FastIntraTools                : 1      # Speedup intra tools: LFNST, ISP, MTS

diff --git a/cfg/randomaccess_medium.cfg b/cfg/randomaccess_medium.cfg
@@ -22,7 +22,7 @@ BipredSearchRange             : 4           # Search range for bi-prediction ref
 HadamardME                    : 1           # Use of hadamard measure for fractional ME
 FEN                           : 3           # Fast encoder decision
 FDM                           : 1           # Fast Decision for Merge RD cost
-NumRefPics                    : 222111      # Number of reference pictures in RPL (0: default for RPL, <10: apply for all temporal layers, >=10: each decimal digit specifies the number for a temporal layer, last digit applying to the highest TL)
+NumRefPics                    : 222221      # Number of reference pictures in RPL (0: default for RPL, <10: apply for all temporal layers, >=10: each decimal digit specifies the number for a temporal layer, last digit applying to the highest TL)
 NumRefPicsSCC                 : 0           # Number of reference pictures in RPL for SCC pictures (semantic analogue to NumRefPics, -1: equal to NumRefPics)
 
 #======== Quantization =============
@@ -58,7 +58,7 @@ DualITree                     : 1      # separate partitioning of luma and chrom
 MinQTLumaISlice               : 8
 MinQTChromaISliceInChromaSamples : 4   # minimum QT size in chroma samples for chroma separate tree
 MinQTNonISlice                : 8
-MaxMTTDepth                   : 221111
+MaxMTTDepth                   : 1
 MaxMTTDepthI                  : 2
 MaxNumMergeCand               : 6      # Maximum number of merge candidates
 
@@ -118,7 +118,7 @@ ContentBasedFastQtbt          : 1      # Signal based QTBT speed-up: 0: disabled
 PBIntraFast                   : 1      # Intra mode pre-check dependent on best Inter mode, skip intra if it is not probable (0:off ... 2:fastest)
 FastQtBtEnc                   : 1      # Fast encoding setting for QTBT
 FastHAD                       : 0      # Use fast sub-sampled hadamard for square blocks >=32x32
-FastMrg                       : 3      # Fast methods for inter merge: 0: disabled, 1: vtm, 2: fast, 3: faster
+FastMrg                       : 2      # Fast methods for inter merge: 0: disabled, 1: vtm, 2: fast, 3: faster
 FastLocalDualTreeMode         : 1      # Fast intra pass coding for local dual-tree in intra coding region: 0: disable, 1: use threshold, 2: one intra mode only
 FastSubPel                    : 1      # Fast sub-pel ME: 0: disabled, 1: enabled
 FastIntraTools                : 1      # Speedup intra tools: LFNST, ISP, MTS

diff --git a/include/vvenc/vvencCfg.h b/include/vvenc/vvencCfg.h
@@ -774,7 +774,7 @@ typedef struct vvenc_config
                                                                                          // if negative, the absolute value is interpreted as a 4-bit fixed point multiplier of the target bitrate).
                                                                                          // -24, i.e. -1.1000 binary, means the maxrate would be set to be the 1.5x of the target bitrate.
                                                                                          // for convenience use VVENC_SET_MAXRATE_FACTOR, e.g. VVENC_SET_MAXRATE_FACTOR(1.5), to set the multiplier
-  int                 m_reservedInt;
+  int                 m_forceScc;
   double              m_reservedDouble[9];
 
   // internal state variables

diff --git a/source/Lib/CommonLib/CommonDef.h b/source/Lib/CommonLib/CommonDef.h
@@ -582,9 +582,11 @@ inline std::string prnt( const char* fmt, ...)
 
 #if ( _WIN32 && ( _MSC_VER > 1300 ) ) || defined (__MINGW64_VERSION_MAJOR)
 #define xMalloc( type, len )        _aligned_malloc( sizeof(type)*(len), MEMORY_ALIGN_DEF_SIZE )
+#define xMalloc2( type, len, alg )  _aligned_malloc( sizeof(type)*(len), alg )
 #define xFree( ptr )                _aligned_free  ( ptr )
 #elif defined (__MINGW32__)
 #define xMalloc( type, len )        __mingw_aligned_malloc( sizeof(type)*(len), MEMORY_ALIGN_DEF_SIZE )
+#define xMalloc2( type, len, alg )  __mingw_aligned_malloc( sizeof(type)*(len), alg )
 #define xFree( ptr )                __mingw_aligned_free( ptr )
 #else
 namespace detail {
@@ -599,11 +601,13 @@ static inline T* aligned_malloc(size_t len, size_t alignement) {
 }
 }
 #define xMalloc( type, len )        detail::aligned_malloc<type>( len, MEMORY_ALIGN_DEF_SIZE )
+#define xMalloc2( type, len, alg )  detail::aligned_malloc<type>( len, alg )
 #define xFree( ptr )                free( ptr )
 #endif
 
 #else
 #define xMalloc( type, len )        malloc   ( sizeof(type)*(len) )
+#define xMalloc2( type, len, alg )  malloc   ( sizeof(type)*(len) )
 #define xFree( ptr )                free     ( ptr )
 #endif //#if ALIGNED_MALLOC