[FFmpeg-devel] AMR-NB decoder

[Diego Biurrun]

On Wed, Aug 05, 2009 at 05:51:36PM +0100, Colin McQuillan wrote:
> Attached is a patch for an AMR-NB decoder.

Build system and documentation part OK.

> --- libavcodec/amrnbdata.h	(revision 0)
> +++ libavcodec/amrnbdata.h	(revision 0)

A lot of stuff could be aligned for better readability here..

> --- libavcodec/amrnbdec.c	(revision 0)
> +++ libavcodec/amrnbdec.c	(revision 0)
> @@ -0,0 +1,1157 @@
> +
> +        const float *gains =
> +            mode >= MODE_67  ? gains_high[amr_subframe->p_gain] :
> +            mode >= MODE_515 ? gains_low [amr_subframe->p_gain] :
> +                // gain index is only coded in subframes 0,2 for MODE_475
> +                gains_MODE_475[(p->frame.info.subframe[subframe & 2].p_gain
> +                                                                       << 1) +
> +                               (subframe & 1)];

This sure looks ugly..

> +        p->pitch_gain[4]  = gains[0];
> +        *fixed_gain_factor = gains[1];

align

> +        // Eqn 67: gc' = 10^0.05 (predicted dB + mean dB - dB of fixed vector)
> +        exp2f(log2f(10.0) * 0.05
> +                        * (ff_dot_productf(energy_pred_fac,
> +                                           p->prediction_error,
> +                                           4) + // predicted fixed energy
> +                           energy_mean[mode])) /

also looks weird

> +        const float **filters = p->cur_frame_mode == MODE_795 ?
> +            ir_filters_lookup_MODE_795 : ir_filters_lookup;

> +        float pitch_factor = (p->cur_frame_mode == MODE_122 ? 0.25 : 0.5)
> +            * FFMIN(p->pitch_gain[4],
> +                    p->cur_frame_mode == MODE_122 ? 1.0 : SHARP_MAX)
> +            * p->pitch_gain[4];

ditto

Diego