I found a bug on A0087 document in HTM reference SW version 8.1 as well as 8.0. In the document, all zero residual method is only applied to inter-picture of depth. However, the method is only applied to intra-picture in both 3D-HTMv8.0 and v8.1.

Line 2995 of TEncSearch.cpp (HTM8.0) is bTestZeroResi |= pcCU->getSlice()->getIsDepth() && pcCU->getSlice()->isIRAP(); It should be fixed to bTestZeroResi |= pcCU->getSlice()->getIsDepth() && !pcCU->getSlice()->isIRAP(); It affects the coding performance of AI and CTC.

In H.8.5.3.3.8,

1. The following equation

xIntL = xP + xL+ disparitySamples[ xL ][ yL ] (H 235) should be modified as xIntL = xP + xL+ (disparitySamples[ xL ][ yL ])>>2 (H 235)

2. The following equation

xIntC = ( xP / 2 ) + xC + disparitySamples[ xC << 1 ][ yC << 1 ] (H 239) should be modified as xIntC = ( xP / 2 ) + xC + (disparitySamples[ xC << 1 ][ yC << 1 ]>>3) (H 239)

Obsolete motion data compression remains in text (I.8.5.5 Derivation process for disparity vectors)

as

1. When yCb − 1 is less than ( ( yCb >> Log2CtbSizeY ) << Log2CtbSizeY ), the following applies:

xB1 = ( ( xB1 >> 3 ) << 3 ) + ( ( xB1 >> 3 ) & 1) * 7 (I 251)

In HTM, the compression have been effectively removed when HTM was updated based on the latest(at that point) HM. Specifically, HTM70 has the following in getPUAbove(),

if(MotionDataCompresssion) {

uiAPartUnitIdx = g_motionRefer[uiAPartUnitIdx];

}

But HTM80 or later doesn't have that process (so removed).