diff options
author | Justin Ruggles <justin.ruggles@gmail.com> | 2012-02-19 17:12:48 -0500 |
---|---|---|
committer | Justin Ruggles <justin.ruggles@gmail.com> | 2012-03-20 18:56:22 -0400 |
commit | b0f75ba272feb465d21cb4520574b8db76c1e954 (patch) | |
tree | b6b7fb3170e4dab17e090b83c19b969b0eeb681a /libavcodec | |
parent | 3d853d7ab317a96a49873f3b3c1848a46f47c7ec (diff) | |
download | ffmpeg-b0f75ba272feb465d21cb4520574b8db76c1e954.tar.gz |
mpegaudioenc: use AVCodec.encode2()
Update FATE references due to encoder delay.
Diffstat (limited to 'libavcodec')
-rw-r--r-- | libavcodec/mpegaudioenc.c | 29 |
1 files changed, 22 insertions, 7 deletions
diff --git a/libavcodec/mpegaudioenc.c b/libavcodec/mpegaudioenc.c index 9ee7f2cba4..385a79a675 100644 --- a/libavcodec/mpegaudioenc.c +++ b/libavcodec/mpegaudioenc.c @@ -80,6 +80,7 @@ static av_cold int MPA_encode_init(AVCodecContext *avctx) bitrate = bitrate / 1000; s->nb_channels = channels; avctx->frame_size = MPA_FRAME_SIZE; + avctx->delay = 512 - 32 + 1; /* encoding freq */ s->lsf = 0; @@ -180,9 +181,11 @@ static av_cold int MPA_encode_init(AVCodecContext *avctx) total_quant_bits[i] = 12 * v; } +#if FF_API_OLD_ENCODE_AUDIO avctx->coded_frame= avcodec_alloc_frame(); if (!avctx->coded_frame) return AVERROR(ENOMEM); +#endif return 0; } @@ -726,14 +729,14 @@ static void encode_frame(MpegAudioContext *s, flush_put_bits(p); } -static int MPA_encode_frame(AVCodecContext *avctx, - unsigned char *frame, int buf_size, void *data) +static int MPA_encode_frame(AVCodecContext *avctx, AVPacket *avpkt, + const AVFrame *frame, int *got_packet_ptr) { MpegAudioContext *s = avctx->priv_data; - const short *samples = data; + const int16_t *samples = (const int16_t *)frame->data[0]; short smr[MPA_MAX_CHANNELS][SBLIMIT]; unsigned char bit_alloc[MPA_MAX_CHANNELS][SBLIMIT]; - int padding, i; + int padding, i, ret; for(i=0;i<s->nb_channels;i++) { filter(s, i, samples + i, s->nb_channels); @@ -748,16 +751,28 @@ static int MPA_encode_frame(AVCodecContext *avctx, } compute_bit_allocation(s, smr, bit_alloc, &padding); - init_put_bits(&s->pb, frame, MPA_MAX_CODED_FRAME_SIZE); + if ((ret = ff_alloc_packet(avpkt, MPA_MAX_CODED_FRAME_SIZE))) { + av_log(avctx, AV_LOG_ERROR, "Error getting output packet\n"); + return ret; + } + + init_put_bits(&s->pb, avpkt->data, avpkt->size); encode_frame(s, bit_alloc, padding); - return put_bits_ptr(&s->pb) - s->pb.buf; + if (frame->pts != AV_NOPTS_VALUE) + avpkt->pts = frame->pts - ff_samples_to_time_base(avctx, avctx->delay); + + avpkt->size = put_bits_count(&s->pb) / 8; + *got_packet_ptr = 1; + return 0; } static av_cold int MPA_encode_close(AVCodecContext *avctx) { +#if FF_API_OLD_ENCODE_AUDIO av_freep(&avctx->coded_frame); +#endif return 0; } @@ -772,7 +787,7 @@ AVCodec ff_mp2_encoder = { .id = CODEC_ID_MP2, .priv_data_size = sizeof(MpegAudioContext), .init = MPA_encode_init, - .encode = MPA_encode_frame, + .encode2 = MPA_encode_frame, .close = MPA_encode_close, .sample_fmts = (const enum AVSampleFormat[]){AV_SAMPLE_FMT_S16,AV_SAMPLE_FMT_NONE}, .supported_samplerates= (const int[]){44100, 48000, 32000, 22050, 24000, 16000, 0}, |