[NONBE] Introduce finite precision comparison in MASA surround coherence at high bitrates
- Related issues:
- Requested reviewers:
Reason why this change is needed
- Starting from a comparison with 0 in BASOP, for which it was hard to reach a sufficient precision, the comparison with 0 is transformed into comparison with a threshold to ensure same precision is used in both float and fixed point versions. The function involved is read_surround_coherence_hr() in MASA for 384kbps and 512kbps.
Description of the change
- the value to be compared with zero is given a finite precision.
Affected operating points
- Produced bitstream is non-BE and non-BC. Old decoder cannot decode the produced bitstream correctly.
- The affected modes are MASA 384kbps and 512kbps and OMASA 384k and 512kbps
Related to #1048 (closed)