Mismatch between Spec and software

[adarsh]

CCLM checks luma reference samples only at the top and left of the current block; this was part of the software from the beginning; it did not check the top-left reference samples. The Spec currently seems to incorrectly refer to the top-left luma reference samples.

The related text Section 8.3.4.2.8 may be referred to using the tag "availTL."

[bbross]

This check is actually needed and thats why I added it.
For the TL sample it is not sufficient to check L and T because with slices (formerly known as tile groups), the case can happen where TL is n.a. but T and L are available or vice versa.
See Figure 6-4 in the current draft where the first CTU in the 6th tile in raster-scan has T and L available but TL is n.a.
Hence, this condition should be added to VTM to cover these cases.

[bbross]

After with Discussion with Karsten, I understand that tiles also inside a slice are independent.
Thus, TL is always n.a. when either L or T is n.a. and as long as we do nat have dependent tiles, the TL available check can be removed from the spec.

[bbross]

Fixed in D5 v6