In TEncCu.cpp line 2935:

m_pcEntropyCoder->encodeQP( rpcTempCU, 0, false );

It should be

m_pcEntropyCoder->encodeQP( rpcTempCU, uiFirstNonZeroPartIdx, false );

When delta QP is on, this bug can cause inaccurate coding of QP in RDO process, thus can affect the RD performance. No impact under CTC.

In the function Void TComAdaptiveLoopFilter::create() in file TComAdaptiveLoopFilter.cpp, line 994: get_mem2Dpel(&m_varImgMethods, m_img_width, m_img_width);

It should be changed as: be get_mem2Dpel(&m_varImgMethods, m_img_height, m_img_width)

Different QPs are assigned to the last I-frame of RAS(N-1) and the first I-frame of RAS(N) where float-point QP is used with parallel processing in JEM4.0. The follows are example of the bug.

• RAS[0], POC 48 TId: 0 ( I-SLICE, nQP 22 QP 22 ) 2871624 bits [Y 48.1091 dB U 44.3101 dB V 44.8896 dB] [ET 5755 ] [L0 ] [L1 ]
• RAS[1], POC 0 TId: 0 ( I-SLICE, nQP 21 QP 21 ) 3232160 bits [Y 48.7618 dB U 44.8700 dB V 45.3831 dB]