[FFmpeg-devel] [PATCH] Add functions shared by AMR and SIPR

[Vitor Sessak]

Vitor Sessak wrote:
> Robert Swain wrote:
>> Vitor Sessak wrote:
>>> Ronald S. Bultje wrote:
>>>> On Sun, Oct 18, 2009 at 7:41 PM, Vitor Sessak <vitor1001 at gmail.com> 
>>>> wrote:
>>>>> The SIPR decoder shares quite a few code with AMR. Most of those 
>>>>> functions
>>>>> are copied from the last AMR patch in -devel. An exception is
>>>>> ff_acelp_interpolatef(), that was copied from a fixed-point version 
>>>>> and is
>>>>> used in AMR in the function interp_pitch_vector().
>>>> [..]
>>>>> +void ff_set_max_dist_lsf(float *lsf, float min_spacing, int size)
>>>>> +{
>>>>> +    int i;
>>>>> +    float prev = 0.0;
>>>>> +    for (i = 0; i < size; i++)
>>>>> +        prev = lsf[i] = FFMAX(lsf[i], prev + min_spacing);
>>>>> +}
>>>>
>>>> ff_set_min_dist_lsf().
>>>
>>> Changed that and Diego's cosmetics in the new version. Michael, ping?
>>
>> Index: libavcodec/acelp_vectors.c
>> ===================================================================
>> --- libavcodec/acelp_vectors.c    (revision 20368)
>> +++ libavcodec/acelp_vectors.c    (working copy)
>>
>> [...]
>>
>> @@ -155,3 +156,22 @@
>>          out[i] = weight_coeff_a * in_a[i]
>>                 + weight_coeff_b * in_b[i];
>>  }
>> +
>> +void ff_adaptative_gain_control(float *buf_out, float speech_energ,
>> +                                int size, float alpha, float *gain_mem)
>> +{
>> +    int i;
>> +    float postfilter_energ = ff_dot_productf(buf_out, buf_out, 48);
>>
>> Why 48? IIRC it should be 40 in AMR-NB. Shouldn't it be replaced with 
>> 'size'?
> 
> Ok, 10l, now actually tested with both AMR and SIPR ;)
> 
> I've attached a modified amrnbdec.c that uses this patch to give an idea 
> of how my patch fit with it. Collin, do you have any opinion on this patch?

I suppose nobody had any more comments, so I've applied it. Please feel 
free to flame me in -cvslogs.

-Vitor