[FFmpeg-devel] [PATCH] ALAC Encoder
Michael Niedermayer
michaelni
Sun Aug 17 04:35:14 CEST 2008
On Sun, Aug 17, 2008 at 04:14:43AM +0530, Jai Menon wrote:
> Hi,
>
> The attached ALAC encoder was written as part of GSoC and mentored by Justin
> Ruggles. I'm posting it for inclusion into FFmpeg-svn.
[...]
> Index: libavcodec/alacenc.c
> ===================================================================
> --- libavcodec/alacenc.c (revision 0)
> +++ libavcodec/alacenc.c (revision 0)
> @@ -0,0 +1,459 @@
> +/**
> + * ALAC audio encoder
> + * Copyright (c) 2008 Jaikrishnan Menon <realityman at gmx.net>
> + *
> + * This file is part of FFmpeg.
> + *
> + * FFmpeg is free software; you can redistribute it and/or
> + * modify it under the terms of the GNU Lesser General Public
> + * License as published by the Free Software Foundation; either
> + * version 2.1 of the License, or (at your option) any later version.
> + *
> + * FFmpeg is distributed in the hope that it will be useful,
> + * but WITHOUT ANY WARRANTY; without even the implied warranty of
> + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
> + * Lesser General Public License for more details.
> + *
> + * You should have received a copy of the GNU Lesser General Public
> + * License along with FFmpeg; if not, write to the Free Software
> + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
> + */
> +
> +#include "avcodec.h"
> +#include "bitstream.h"
> +#include "dsputil.h"
> +#include "lpc.h"
> +
> +#define DEFAULT_FRAME_SIZE 4096
> +#define DEFAULT_SAMPLE_SIZE 16
> +#define MAX_CHANNELS 8
> +#define ALAC_EXTRADATA_SIZE 36
> +#define ALAC_FRAME_HEADER_SIZE 55
> +#define ALAC_FRAME_FOOTER_SIZE 3
> +
> +#define ALAC_ESCAPE_CODE 0x1FF
> +#define ALAC_MAX_LPC_ORDER 30
ok
[...]
> +typedef struct AlacEncodeContext {
> + int channels;
> + int samplerate;
redundant relative to the fields in AVCodecContext
[...]
> + int interlacing_shift;
> + int interlacing_leftweight;
> + PutBitContext pbctx;
ok (pb is more common than pbctx but this is nitpicking, pbctx is ok if you
prefer, the same is true for dspctx)
[...]
> + DSPContext dspctx;
> + AVCodecContext *avctx;
> +} AlacEncodeContext;
ok
> +
> +
> +static void allocate_sample_buffers(AlacEncodeContext *s)
> +{
> + int i = s->channels;
> +
> + while(i) {
> + s->sample_buf[i-1] = av_mallocz(s->avctx->frame_size*sizeof(int32_t));
> + i--;
> + }
> + s->predictor_buf = av_mallocz(s->avctx->frame_size*sizeof(int32_t));
> +}
> +
> +static void free_sample_buffers(AlacEncodeContext *s)
> +{
> + int i = s->channels;
> +
> + while(i) {
> + av_freep(&s->sample_buf[i-1]);
> + i--;
> + }
> + av_freep(&s->predictor_buf);
> +}
As they have constant size they do not need a malloc() but instead can be part
of the context
[...]
> +static void encode_scalar(AlacEncodeContext *s, int x, int k, int write_sample_size)
> +{
> + int divisor, q, r;
> +
> + k = FFMIN(k, s->rc.k_modifier);
> + divisor = (1<<k) - 1;
> + q = x / divisor;
> + r = x % divisor;
> +
> + if(q > 8) {
> + // write escape code and sample value directly
> + put_bits(&s->pbctx, 9, ALAC_ESCAPE_CODE);
> + put_bits(&s->pbctx, write_sample_size, x);
> + } else {
> + if(q)
> + put_bits(&s->pbctx, q, (1<<q) - 1);
> + put_bits(&s->pbctx, 1, 0);
> +
> + if(k != 1) {
> + if(r > 0)
> + put_bits(&s->pbctx, k, r+1);
> + else
> + put_bits(&s->pbctx, k-1, 0);
> + }
> + }
> +}
ok
> +
> +static void write_frame_header(AlacEncodeContext *s, int is_verbatim)
> +{
> + put_bits(&s->pbctx, 3, s->channels-1); // No. of channels -1
> + put_bits(&s->pbctx, 16, 0); // Seems to be zero
> + put_bits(&s->pbctx, 1, 1); // Sample count is in the header
> + put_bits(&s->pbctx, 2, 0); // FIXME: Wasted bytes field
> + put_bits(&s->pbctx, 1, is_verbatim); // Audio block is verbatim
> + put_bits(&s->pbctx, 32, s->avctx->frame_size); // No. of samples in the frame
> +}
ok
[...]
> +static void alac_stereo_decorrelation(AlacEncodeContext *s)
> +{
> + int32_t *left = s->sample_buf[0], *right = s->sample_buf[1];
> + int32_t tmp;
> + int i;
> +
> + for(i=0; i<s->avctx->frame_size; i++) {
> + tmp = left[i];
> + left[i] = (tmp + right[i]) >> 1;
> + right[i] = tmp - right[i];
> + }
> + s->interlacing_leftweight = 1;
> + s->interlacing_shift = 1;
i do not belive this is optimal
> +}
> +
> +static void alac_linear_predictor(AlacEncodeContext *s, int ch)
> +{
> + int i;
> + LPCContext lpc = s->lpc[ch];
> +
> + if(lpc.lpc_order == 31) {
> + s->predictor_buf[0] = s->sample_buf[ch][0];
> + i = s->avctx->frame_size - 1;
> + while(i > 0) {
> + s->predictor_buf[i] = s->sample_buf[ch][i] - s->sample_buf[ch][i-1];
> + i--;
> + }
i suspect that 0 .. n is faster than n .. 0
[...]
> +static void alac_entropy_coder(AlacEncodeContext *s)
> +{
> + unsigned int history = s->rc.initial_history;
> + int sign_modifier = 0, i = 0, k;
> + int32_t *samples = s->predictor_buf;
> +
> + while(i < s->avctx->frame_size) {
for(i=0; i<s->avctx->frame_size;
[...]
> +static void write_compressed_frame(AlacEncodeContext *s)
> +{
> + int i, j;
> +
> + /* only simple mid/side decorrelation supported as of now */
> + alac_stereo_decorrelation(s);
> + put_bits(&s->pbctx, 8, s->interlacing_shift);
> + put_bits(&s->pbctx, 8, s->interlacing_leftweight);
> +
> + for(i=0;i<s->channels;i++) {
> +
> + 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, s->lpc[i].lpc_quant);
> +
> + put_bits(&s->pbctx, 3, s->rc.rice_modifier);
> + put_bits(&s->pbctx, 5, s->lpc[i].lpc_order);
> + // predictor coeff. table
> + for(j=0;j<s->lpc[i].lpc_order;j++) {
> + put_sbits(&s->pbctx, 16, s->lpc[i].lpc_coeff[j]);
> + }
> + }
> +
> + // apply lpc and entropy coding to audio samples
> +
> + for(i=0;i<s->channels;i++) {
> + alac_linear_predictor(s, i);
> + alac_entropy_coder(s);
> + }
> +}
ok
> +static av_cold int alac_encode_init(AVCodecContext *avctx)
> +{
> + AlacEncodeContext *s = avctx->priv_data;
> + uint8_t *alac_extradata = av_mallocz(ALAC_EXTRADATA_SIZE+1);
> +
> + avctx->frame_size = DEFAULT_FRAME_SIZE;
> + avctx->bits_per_sample = DEFAULT_SAMPLE_SIZE;
> + s->channels = avctx->channels;
> + s->samplerate = avctx->sample_rate;
> +
> + if(avctx->sample_fmt != SAMPLE_FMT_S16) {
> + av_log(avctx, AV_LOG_ERROR, "only pcm_s16 input samples are supported\n");
> + return -1;
> + }
> +
> + // Set default compression level
> + if(avctx->compression_level == FF_COMPRESSION_DEFAULT)
> + s->compression_level = 1;
> + else
> + s->compression_level = av_clip(avctx->compression_level, 0, 1);
> +
> + // Initialize default Rice parameters
> + s->rc.history_mult = 40;
> + s->rc.initial_history = 10;
> + s->rc.k_modifier = 14;
> + s->rc.rice_modifier = 4;
> +
> + s->max_coded_frame_size = (ALAC_FRAME_HEADER_SIZE + ALAC_FRAME_FOOTER_SIZE +
> + avctx->frame_size*s->channels*avctx->bits_per_sample)>>3;
> +
> + s->write_sample_size = avctx->bits_per_sample + s->channels - 1; // FIXME: consider wasted_bytes
> +
> + AV_WB32(alac_extradata, ALAC_EXTRADATA_SIZE);
> + AV_WB32(alac_extradata+4, MKBETAG('a','l','a','c'));
> + AV_WB32(alac_extradata+12, avctx->frame_size);
> + AV_WB8 (alac_extradata+17, avctx->bits_per_sample);
> + AV_WB8 (alac_extradata+21, s->channels);
> + AV_WB32(alac_extradata+24, s->max_coded_frame_size);
> + AV_WB32(alac_extradata+28, s->samplerate*s->channels*avctx->bits_per_sample); // average bitrate
> + AV_WB32(alac_extradata+32, s->samplerate);
> +
> + // Set relevant extradata fields
> + if(s->compression_level > 0) {
> + AV_WB8(alac_extradata+18, s->rc.history_mult);
> + AV_WB8(alac_extradata+19, s->rc.initial_history);
> + AV_WB8(alac_extradata+20, s->rc.k_modifier);
> + }
> +
> + avctx->extradata = alac_extradata;
> + avctx->extradata_size = ALAC_EXTRADATA_SIZE;
> +
> + avctx->coded_frame = avcodec_alloc_frame();
> + avctx->coded_frame->key_frame = 1;
> +
> + s->avctx = avctx;
> + dsputil_init(&s->dspctx, avctx);
> +
> + allocate_sample_buffers(s);
> +
> + return 0;
> +}
ok
> +
> +static int alac_encode_frame(AVCodecContext *avctx, uint8_t *frame,
> + int buf_size, void *data)
> +{
> + AlacEncodeContext *s = avctx->priv_data;
> + PutBitContext *pb = &s->pbctx;
> + int i, out_bytes;
> +
> + if(avctx->frame_size > DEFAULT_FRAME_SIZE) {
> + av_log(avctx, AV_LOG_ERROR, "input frame size exceeded\n");
> + return -1;
> + }
> +
> + if(buf_size < s->max_coded_frame_size) {
> + av_log(avctx, AV_LOG_ERROR, "buffer size is too small\n");
> + return -1;
> + }
> +
> + init_put_bits(pb, frame, buf_size);
> +
> + if(s->compression_level == 0) {
> + // Verbatim mode
> + int16_t *samples = data;
> + write_frame_header(s, 1);
> + for(i=0; i<avctx->frame_size*s->channels; i++) {
> + put_sbits(pb, 16, *samples++);
> + }
> + } else {
> + init_sample_buffers(s, data);
> + write_frame_header(s, 0);
> + write_compressed_frame(s);
> + }
> +
> + put_bits(pb, 3, 7);
> + flush_put_bits(pb);
> + out_bytes = put_bits_count(pb) >> 3;
> +
> + if(out_bytes > s->max_coded_frame_size) {
At this point it is possible that the encoder wrote over the end of the
array, the check farther up should ensure that there really is enough
space available even for larger then optimal frames
> + /* frame too large. use verbatim mode */
> + int16_t *samples = data;
> + init_put_bits(pb, frame, buf_size);
> + write_frame_header(s, 0);
is_verbatim = 0 ?
> +
> + for(i=0; i<avctx->frame_size*s->channels; i++) {
> + put_sbits(pb, 16, *samples++);
> + }
> + put_bits(pb, 3, 7);
> + flush_put_bits(pb);
> + out_bytes = put_bits_count(pb) >> 3;
all this looks a little duplicated and could be simplified by a goto
> +
> + if(out_bytes > s->max_coded_frame_size || out_bytes >= buf_size) {
> + /* still too large. must be an error. */
> + av_log(avctx, AV_LOG_ERROR, "error encoding frame\n");
> + return -1;
> + }
> + }
> +
> + return out_bytes;
> +}
> +
> +static av_cold int alac_encode_close(AVCodecContext *avctx)
> +{
> + AlacEncodeContext *s = avctx->priv_data;
> +
> + av_freep(&avctx->extradata);
> + avctx->extradata_size = 0;
> + av_freep(&avctx->coded_frame);
> + free_sample_buffers(s);
> + return 0;
> +}
> +
> +AVCodec alac_encoder = {
> + "alac",
> + CODEC_TYPE_AUDIO,
> + CODEC_ID_ALAC,
> + sizeof(AlacEncodeContext),
> + alac_encode_init,
> + alac_encode_frame,
> + alac_encode_close,
> + .capabilities = CODEC_CAP_SMALL_LAST_FRAME,
> + .long_name = "ALAC (Apple Lossless Audio Codec)",
> +};
ok
[...]
--
Michael GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB
Let us carefully observe those good qualities wherein our enemies excel us
and endeavor to excel them, by avoiding what is faulty, and imitating what
is excellent in them. -- Plutarch
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: Digital signature
URL: <http://lists.mplayerhq.hu/pipermail/ffmpeg-devel/attachments/20080817/083f30c0/attachment.pgp>
More information about the ffmpeg-devel
mailing list