basop-2338-harmonise-duplication-of-comp_and_apply_gain_fx[non-BE] [allow-regression] [split-non-BE]
- Related issues: #2338
- Requested reviewers:
Reason why this change is needed
- There are 2 functions
Comp_and_apply_gain_fx()andComp_and_apply_gain_IVAS_fx()(related to the GSC ACELP coding) that do basically the same, except that the IVAS version truncates the excitation vector before performing a final upscaling based on the minimum value of the truncated vector. The EVS version used a maybe less optimal final scaling value, but it doesn't truncate. For the vast majority of cases, the EVS version will be more precise..
Description of the change
- Simply replacing the occurence of
Comp_and_apply_gain_IVAS_fxwithComp_and_apply_gain_fx
Affected operating points
- All operating points where GSC is used in IVAS
- Bitstream and synthesis not bit-exact, but inter-operable
Edited by vaillancour