Custom Query (1442 matches)
Results (19 - 21 of 1442)
Ticket | Resolution | Summary | Owner | Reporter |
---|---|---|---|---|
#19 | fixed | Memory leak in TEncAdaptiveLoopFilter.cpp (TMuC 0.5, revision 81) | li@… | |
Description |
Memory leaks are observed in Void TEncAdaptiveLoopFilter::startALFEnc( TComPic* pcPic, TEncEntropy* pcEntropyCoder ) (HHI_ALF off) Member variables m_pcPicYuvTmp, m_pcBestAlfParam, m_pcTempAlfParam, ALFp, tempALFp are allocated in this function at frame level. But they are never released. The problem becomes serious when many frames are coded. A patch of bug fix is attached. |
|||
#20 | fixed | Turning off QC_MDDT macro makes compile error (rev 67) | hurumi@… | |
Description |
Turning off QC_MDDT macro makes compile-time error like: 1>..\..\source\Lib\TLibCommon\TComTrQuant.cpp(4672) : error C2065: 'uiMode' : undeclared identifier 1>..\..\source\Lib\TLibCommon\TComTrQuant.cpp(4672) : error C3861: 'getUseMDDT': identifier not found 1>..\..\source\Lib\TLibCommon\TComTrQuant.cpp(4780) : error C2065: 'uiMode' : undeclared identifier 1>..\..\source\Lib\TLibCommon\TComTrQuant.cpp(4780) : error C3861: 'getUseMDDT': identifier not found It seems to be related to the NEWVLC macro which uses some functions defined only when QC_MDDT is equal to 1. |
|||
#21 | fixed | Frequent memory allocation (TMuC 0.5, revision 81) | fbossen | li@… |
Description |
During tracing memory leak, we noticed a problem with the current ALF, CABAC and PIPE. When using encoder_lowdelay_loco.cfg code 3 WQGVA frames (gop size=2, 1 intra frame), 4452 times of memory allocation were recorded. When using encoder_lowdelay_loco.cfg and --ALF=1, memory allocation is increased to 327,753 times. When using encoder_lowdely_loco.cfg and --SymbolMode=1 (CABAC), memory allocation is increased to 12,396,839 times. When using encoder_lowdely_loco.cfg and --SymbolMode=2 (PIPE), memory allocation is increased to 12,396,839 times. (I was some surprised that we got the exact same number for CABAC and PIPE.) In principle, too frequent memory allocation will slow down the coding speed and lead to memory fragmentation. We think it is better to improve the related code. |