id	summary	reporter	owner	description	type	status	priority	milestone	component	version	resolution	keywords	cc
180	HM-3.1 bug: Previous QP is not correct	hao		"According to the decision at the Geneva meeting, new QP predictor for WD3 is introduced into HM-3.1, with the macro SUB_LCU_DQP. The predictor is defined as follows.
1. Use left negihbor (of upper-left corner of currentCU) if available
2. Otherwise, use previous in scan order.

However, the current implementation fails to correctly get previous QP in scan order.
TComDataCU::m_hLastCodedQP representing the previous QP is not set to the real previous QP but set to the reference QP when encoding/decoding the current QP. As a result, correct predicted QP cannot be obtained at the first CU in the left-most LCU.
I believe the correct procedure is that TComDataCU::m_hLastCodedQP is set to the current QP when finishing encoding/decoding process of CU's whose depth is equal to or less than MaxCuDQPDepth.

In addition, QP of CU's coded as PCM is not correctly set."	defect	closed	critical		HM		fixed		fbossen ksuehring davidf jct-vc@…