At Line 391 @ TComLoopFilter.cpp, the value of parameter "iBaseUnitIdx" should be set as "const UInt iBaseUnitIdx = uiLCUWidthInBaseUnits >> ( uiDepth + 2 );" instead of "const UInt iBaseUnitIdx = uiLCUWidthInBaseUnits >> ( uiDepth + 1 );". Considering that non-square transform block will be split to four smaller blocks with half horizontal and vertical size at this condition, the parameter "iBaseUnitIdx" is set to a wrong value at HM4.0.

This bug doesn't influence all the testing cases under common test condition. But it could cause mismatch if "QuadtreeTUMaxDepthInter" is set to 1.

Compared to HM4.0-dev-1354 version, the performance of the bug fixed version under common test condition is as follow:

	Random Access HE
	Y	U	V
Class A	0.0%	-0.2%	-0.1%
Class B	0.0%	0.0%	0.1%
Class C	0.0%	0.0%	0.1%
Class D	0.0%	0.0%	0.0%
Class E
Class F	-0.1%	-0.1%	-0.1%
Overall	0.0%	0.0%	0.0%
	0.0%	0.0%	0.0%
Enc Time[%]	100%
Dec Time[%]	99%

	Random Access LC
	Y	U	V
Class A	0.0%	0.0%	0.0%
Class B	0.0%	0.0%	0.1%
Class C	0.0%	0.0%	0.0%
Class D	0.0%	0.0%	0.0%
Class E
Class F	0.0%	-0.1%	0.0%
Overall	0.0%	0.0%	0.0%
	0.0%	0.0%	0.0%
Enc Time[%]	100%
Dec Time[%]	100%

	Low delay B HE
	Y	U	V
Class A
Class B	0.0%	-0.1%	-0.2%
Class C	0.0%	0.1%	-0.1%
Class D	0.0%	0.5%	-0.1%
Class E	-0.1%	0.0%	0.1%
Class F	-0.1%	-0.1%	0.0%
Overall	0.0%	0.1%	-0.1%
	0.0%	-0.1%	0.0%
Enc Time[%]	100%
Dec Time[%]	98%

	Low delay B HE
	Y	U	V
Class A
Class B	0.0%	0.1%	0.2%
Class C	0.0%	0.1%	-0.2%
Class D	0.0%	0.0%	-0.2%
Class E	-0.1%	-0.1%	-0.1% |
Class F	0.1%	0.3%	0.0%
Overall	0.0%	0.1%	-0.1%
	0.0%	-0.1%	0.0%
Enc Time[%]	100%
Dec Time[%]	98%