max_vps_dec_pic_buffering_minus1[ i ][ k ][ j ] plus 1, when NecessaryLayerFlag[ i ][ k ] is equal to 1, specifies the maximum number of decoded pictures, of the k-th layer for the CVS in the i-th OLS, that need to be stored in the DPB when HighestTid is equal to j.

Semantics is ambiguous when saying "k-th layer for the CVS in the i-th OLS". It can be understood as a layer with index k in CVS in the i-th OLS, however it is a layer with index k, where the k is the index within layers included into i-th OLS.

Probably, "for the CVS" can be removed or moved after OLS.

There are the following constraints defined in the semantics of inter_layer_pred_layer_idc[i].

It is a requirement of bitstream conformance that for each value of i in the range of 0 to NumActiveRefLayerPics − 1, inclusive, either of the following two conditions shall be true:

• The value of max_tid_il_ref_pics_plus1[ LayerIdxInVps[ RefPicLayerId[ i ] ] ][ LayerIdxInVps[ nuh_layer_id ] ] is greater than TemporalId.

• The values of max_tid_il_ref_pics_plus1[ LayerIdxInVps[ RefPicLayerId[ i ] ] ][ LayerIdxInVps[ nuh_layer_id ] ] and TemporalId are both equal to 0 and the picture in the current access unit with nuh_layer_id equal to RefPicLayerId[ i ] is an IRAP picture.

It seems that highlighted temporal Id should be temporal Id of the reference layer picture rather than the current picture TemporalId.

When using multiple slices per picture, such as SlideMode=2 (other settings are the same as SHVC CTC), decoding crash is observed with SHM-3.x software.