|
- 1. All input contributions were discussed with the following conclusions:
Inverse Quantization:
the iq_mode can come from a provider or from the bitstream
Inverse Binarization is a normal FU
Decode Geometry: add "TFAN" to the name since it is tfan specific (accepted).
TFAN Network: update the CE to check if in RVC - mngmt FUs can be used
CD:
- create a section called "Examples of Decoder networks using GTL": TFAN (code, picture), SVA (code, picture), BBA (code, picture).
- Editing period: to check when a FU should loop back at the START state
- editorial period : -> IDCT: - modify it to load the matrix from a provider
- define segment_size (how many tokens are processed using the same matrix)
-> Changes expected in LUTs and providers depending on the results of the FNL study
- CE description: Study if the inverse circular prediction is the same process as in IP FU / all FUs should be described as pseudo code (not C like code)
doc 25077:
Editorial period:
-> VLD decoder: check copy-past error on input table
- also, be more clear in the process description pseudo code (store 2D array ?)
-> Review the table with FU names (to correspond with the description)
-> IDCT decoder: check copy-past error on input table
- also, be more clear in the process description pseudo code (store 2D array ?)
-> Matrix2Vector decoder: check copy-past error on input table
- also, be more clear in the process description pseudo code (store 2D array ?)
-> Generate Alphabet:
- check pseudo code (array as input/output but the code refers the non-array current token)
- Explain the distribution var more detailed
Token types:
-> consider to provide a table as an informative note in the CD that describes the possible token types and combinations for each FU - accepted
Frame Work:
-> Output document describing how to install and use the frameworks.
-> Issue RefSof and conformance AMD
| 
