alacenc: implement the 2-pass prediction type.

This isn't used by the reference encoder, but it is supported by the decoder.
This commit is contained in:
Justin Ruggles 2011-12-07 13:01:01 -05:00
parent 149e1b0468
commit bb63475aec
1 changed files with 10 additions and 1 deletions

View File

@ -348,6 +348,7 @@ static void alac_entropy_coder(AlacEncodeContext *s)
static void write_compressed_frame(AlacEncodeContext *s)
{
int i, j;
int prediction_type = 0;
if (s->avctx->channels == 2)
alac_stereo_decorrelation(s);
@ -358,7 +359,7 @@ static void write_compressed_frame(AlacEncodeContext *s)
calc_predictor_params(s, i);
put_bits(&s->pbctx, 4, 0); // prediction type : currently only type 0 has been RE'd
put_bits(&s->pbctx, 4, prediction_type);
put_bits(&s->pbctx, 4, s->lpc[i].lpc_quant);
put_bits(&s->pbctx, 3, s->rc.rice_modifier);
@ -373,6 +374,14 @@ static void write_compressed_frame(AlacEncodeContext *s)
for (i = 0; i < s->avctx->channels; i++) {
alac_linear_predictor(s, i);
// TODO: determine when this will actually help. for now it's not used.
if (prediction_type == 15) {
// 2nd pass 1st order filter
for (j = s->avctx->frame_size - 1; j > 0; j--)
s->predictor_buf[j] -= s->predictor_buf[j - 1];
}
alac_entropy_coder(s);
}
}