#include "avcodec.h"
#include "put_bits.h"
#include "dsputil.h"
+#include "internal.h"
#include "lpc.h"
#include "mathops.h"
} AlacLPCContext;
typedef struct AlacEncodeContext {
+ int frame_size; /**< current frame size */
+ int verbatim; /**< current frame verbatim mode flag */
int compression_level;
int min_prediction_order;
int max_prediction_order;
} AlacEncodeContext;
-static void init_sample_buffers(AlacEncodeContext *s,
- const int16_t *input_samples)
+static void init_sample_buffers(AlacEncodeContext *s, int16_t **input_samples)
{
int ch, i;
for (ch = 0; ch < s->avctx->channels; ch++) {
- const int16_t *sptr = input_samples + ch;
- for (i = 0; i < s->avctx->frame_size; i++) {
- s->sample_buf[ch][i] = *sptr;
- sptr += s->avctx->channels;
- }
+ int32_t *bptr = s->sample_buf[ch];
+ const int16_t *sptr = input_samples[ch];
+ for (i = 0; i < s->frame_size; i++)
+ bptr[i] = sptr[i];
}
}
}
}
-static void write_frame_header(AlacEncodeContext *s, int is_verbatim)
+static void write_frame_header(AlacEncodeContext *s)
{
- put_bits(&s->pbctx, 3, s->avctx->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_bits32(&s->pbctx, s->avctx->frame_size); // No. of samples in the frame
+ int encode_fs = 0;
+
+ if (s->frame_size < DEFAULT_FRAME_SIZE)
+ encode_fs = 1;
+
+ put_bits(&s->pbctx, 3, s->avctx->channels-1); // No. of channels -1
+ put_bits(&s->pbctx, 16, 0); // Seems to be zero
+ put_bits(&s->pbctx, 1, encode_fs); // Sample count is in the header
+ put_bits(&s->pbctx, 2, 0); // FIXME: Wasted bytes field
+ put_bits(&s->pbctx, 1, s->verbatim); // Audio block is verbatim
+ if (encode_fs)
+ put_bits32(&s->pbctx, s->frame_size); // No. of samples in the frame
}
static void calc_predictor_params(AlacEncodeContext *s, int ch)
s->lpc[ch].lpc_coeff[5] = -25;
} else {
opt_order = ff_lpc_calc_coefs(&s->lpc_ctx, s->sample_buf[ch],
- s->avctx->frame_size,
+ s->frame_size,
s->min_prediction_order,
s->max_prediction_order,
ALAC_MAX_LPC_PRECISION, coefs, shift,
/* calculate sum of 2nd order residual for each channel */
sum[0] = sum[1] = sum[2] = sum[3] = 0;
for (i = 2; i < n; i++) {
- lt = left_ch[i] - 2*left_ch[i-1] + left_ch[i-2];
- rt = right_ch[i] - 2*right_ch[i-1] + right_ch[i-2];
+ lt = left_ch[i] - 2 * left_ch[i - 1] + left_ch[i - 2];
+ rt = right_ch[i] - 2 * right_ch[i - 1] + right_ch[i - 2];
sum[2] += FFABS((lt + rt) >> 1);
sum[3] += FFABS(lt - rt);
sum[0] += FFABS(lt);
/* return mode with lowest score */
best = 0;
for (i = 1; i < 4; i++) {
- if (score[i] < score[best]) {
+ if (score[i] < score[best])
best = i;
- }
}
return best;
}
static void alac_stereo_decorrelation(AlacEncodeContext *s)
{
int32_t *left = s->sample_buf[0], *right = s->sample_buf[1];
- int i, mode, n = s->avctx->frame_size;
+ int i, mode, n = s->frame_size;
int32_t tmp;
mode = estimate_stereo_mode(left, right, n);
- switch(mode)
- {
- case ALAC_CHMODE_LEFT_RIGHT:
- s->interlacing_leftweight = 0;
- s->interlacing_shift = 0;
- break;
-
- case ALAC_CHMODE_LEFT_SIDE:
- for (i = 0; i < n; i++) {
- right[i] = left[i] - right[i];
- }
- s->interlacing_leftweight = 1;
- s->interlacing_shift = 0;
- break;
-
- case ALAC_CHMODE_RIGHT_SIDE:
- for (i = 0; i < n; i++) {
- tmp = right[i];
- right[i] = left[i] - right[i];
- left[i] = tmp + (right[i] >> 31);
- }
- s->interlacing_leftweight = 1;
- s->interlacing_shift = 31;
- break;
-
- default:
- for (i = 0; i < n; i++) {
- tmp = left[i];
- left[i] = (tmp + right[i]) >> 1;
- right[i] = tmp - right[i];
- }
- s->interlacing_leftweight = 1;
- s->interlacing_shift = 1;
- break;
+ switch (mode) {
+ case ALAC_CHMODE_LEFT_RIGHT:
+ s->interlacing_leftweight = 0;
+ s->interlacing_shift = 0;
+ break;
+ case ALAC_CHMODE_LEFT_SIDE:
+ for (i = 0; i < n; i++)
+ right[i] = left[i] - right[i];
+ s->interlacing_leftweight = 1;
+ s->interlacing_shift = 0;
+ break;
+ case ALAC_CHMODE_RIGHT_SIDE:
+ for (i = 0; i < n; i++) {
+ tmp = right[i];
+ right[i] = left[i] - right[i];
+ left[i] = tmp + (right[i] >> 31);
+ }
+ s->interlacing_leftweight = 1;
+ s->interlacing_shift = 31;
+ break;
+ default:
+ for (i = 0; i < n; i++) {
+ tmp = left[i];
+ left[i] = (tmp + right[i]) >> 1;
+ right[i] = tmp - right[i];
+ }
+ s->interlacing_leftweight = 1;
+ s->interlacing_shift = 1;
+ break;
}
}
if (lpc.lpc_order == 31) {
s->predictor_buf[0] = s->sample_buf[ch][0];
- for (i = 1; i < s->avctx->frame_size; i++)
- s->predictor_buf[i] = s->sample_buf[ch][i] - s->sample_buf[ch][i-1];
+ for (i = 1; i < s->frame_size; i++) {
+ s->predictor_buf[i] = s->sample_buf[ch][i ] -
+ s->sample_buf[ch][i - 1];
+ }
return;
}
residual[i] = samples[i] - samples[i-1];
// perform lpc on remaining samples
- for (i = lpc.lpc_order + 1; i < s->avctx->frame_size; i++) {
+ for (i = lpc.lpc_order + 1; i < s->frame_size; i++) {
int sum = 1 << (lpc.lpc_quant - 1), res_val, j;
for (j = 0; j < lpc.lpc_order; j++) {
sum += (samples[lpc.lpc_order-j] - samples[0]) *
- lpc.lpc_coeff[j];
+ lpc.lpc_coeff[j];
}
sum >>= lpc.lpc_quant;
s->write_sample_size);
res_val = residual[i];
- if(res_val) {
+ if (res_val) {
int index = lpc.lpc_order - 1;
int neg = (res_val < 0);
- while(index >= 0 && (neg ? (res_val < 0):(res_val > 0))) {
- int val = samples[0] - samples[lpc.lpc_order - index];
+ while (index >= 0 && (neg ? (res_val < 0) : (res_val > 0))) {
+ int val = samples[0] - samples[lpc.lpc_order - index];
int sign = (val ? FFSIGN(val) : 0);
- if(neg)
- sign*=-1;
+ if (neg)
+ sign *= -1;
lpc.lpc_coeff[index] -= sign;
val *= sign;
- res_val -= ((val >> lpc.lpc_quant) *
- (lpc.lpc_order - index));
+ res_val -= (val >> lpc.lpc_quant) * (lpc.lpc_order - index);
index--;
}
}
int sign_modifier = 0, i, k;
int32_t *samples = s->predictor_buf;
- for (i = 0; i < s->avctx->frame_size;) {
+ for (i = 0; i < s->frame_size;) {
int x;
k = av_log2((history >> 9) + 3);
- x = -2*(*samples)-1;
- x ^= (x>>31);
+ x = -2 * (*samples) -1;
+ x ^= x >> 31;
samples++;
i++;
encode_scalar(s, x - sign_modifier, k, s->write_sample_size);
- history += x * s->rc.history_mult
- - ((history * s->rc.history_mult) >> 9);
+ history += x * s->rc.history_mult -
+ ((history * s->rc.history_mult) >> 9);
sign_modifier = 0;
if (x > 0xFFFF)
history = 0xFFFF;
- if (history < 128 && i < s->avctx->frame_size) {
+ if (history < 128 && i < s->frame_size) {
unsigned int block_size = 0;
k = 7 - av_log2(history) + ((history + 16) >> 6);
- while (*samples == 0 && i < s->avctx->frame_size) {
+ while (*samples == 0 && i < s->frame_size) {
samples++;
i++;
block_size++;
}
encode_scalar(s, block_size, k, 16);
-
sign_modifier = (block_size <= 0xFFFF);
-
history = 0;
}
}
}
-static void write_compressed_frame(AlacEncodeContext *s)
+static int write_frame(AlacEncodeContext *s, AVPacket *avpkt, int16_t **samples)
{
int i, j;
int prediction_type = 0;
+ PutBitContext *pb = &s->pbctx;
- if (s->avctx->channels == 2)
- alac_stereo_decorrelation(s);
- put_bits(&s->pbctx, 8, s->interlacing_shift);
- put_bits(&s->pbctx, 8, s->interlacing_leftweight);
+ init_put_bits(pb, avpkt->data, avpkt->size);
+
+ if (s->verbatim) {
+ write_frame_header(s);
+ /* samples are channel-interleaved in verbatim mode */
+ for (i = 0; i < s->frame_size; i++)
+ for (j = 0; j < s->avctx->channels; j++)
+ put_sbits(pb, 16, samples[j][i]);
+ } else {
+ init_sample_buffers(s, samples);
+ write_frame_header(s);
- for (i = 0; i < s->avctx->channels; i++) {
+ if (s->avctx->channels == 2)
+ alac_stereo_decorrelation(s);
+ put_bits(pb, 8, s->interlacing_shift);
+ put_bits(pb, 8, s->interlacing_leftweight);
- calc_predictor_params(s, i);
+ for (i = 0; i < s->avctx->channels; i++) {
+ calc_predictor_params(s, i);
- put_bits(&s->pbctx, 4, prediction_type);
- put_bits(&s->pbctx, 4, s->lpc[i].lpc_quant);
+ put_bits(pb, 4, prediction_type);
+ put_bits(pb, 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]);
+ put_bits(pb, 3, s->rc.rice_modifier);
+ put_bits(pb, 5, s->lpc[i].lpc_order);
+ // predictor coeff. table
+ for (j = 0; j < s->lpc[i].lpc_order; j++)
+ put_sbits(pb, 16, s->lpc[i].lpc_coeff[j]);
}
- }
- // apply lpc and entropy coding to audio samples
+ // apply lpc and entropy coding to audio samples
- for (i = 0; i < s->avctx->channels; i++) {
- alac_linear_predictor(s, i);
+ 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];
- }
+ // 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->frame_size - 1; j > 0; j--)
+ s->predictor_buf[j] -= s->predictor_buf[j - 1];
+ }
- alac_entropy_coder(s);
+ alac_entropy_coder(s);
+ }
}
+ put_bits(pb, 3, 7);
+ flush_put_bits(pb);
+ return put_bits_count(pb) >> 3;
+}
+
+static av_always_inline int get_max_frame_size(int frame_size, int ch, int bps)
+{
+ int header_bits = 23 + 32 * (frame_size < DEFAULT_FRAME_SIZE);
+ return FFALIGN(header_bits + bps * ch * frame_size + 3, 8) / 8;
}
static av_cold int alac_encode_close(AVCodecContext *avctx)
static av_cold int alac_encode_init(AVCodecContext *avctx)
{
- AlacEncodeContext *s = avctx->priv_data;
+ AlacEncodeContext *s = avctx->priv_data;
int ret;
uint8_t *alac_extradata;
- avctx->frame_size = DEFAULT_FRAME_SIZE;
-
- if (avctx->sample_fmt != AV_SAMPLE_FMT_S16) {
- av_log(avctx, AV_LOG_ERROR, "only pcm_s16 input samples are supported\n");
- return -1;
- }
+ avctx->frame_size = s->frame_size = DEFAULT_FRAME_SIZE;
/* TODO: Correctly implement multi-channel ALAC.
It is similar to multi-channel AAC, in that it has a series of
s->rc.k_modifier = 14;
s->rc.rice_modifier = 4;
- s->max_coded_frame_size = 8 + (avctx->frame_size * avctx->channels * DEFAULT_SAMPLE_SIZE >> 3);
+ s->max_coded_frame_size = get_max_frame_size(avctx->frame_size,
+ avctx->channels,
+ DEFAULT_SAMPLE_SIZE);
- s->write_sample_size = DEFAULT_SAMPLE_SIZE + avctx->channels - 1; // FIXME: consider wasted_bytes
+ // FIXME: consider wasted_bytes
+ s->write_sample_size = DEFAULT_SAMPLE_SIZE + avctx->channels - 1;
avctx->extradata = av_mallocz(ALAC_EXTRADATA_SIZE + FF_INPUT_BUFFER_PADDING_SIZE);
if (!avctx->extradata) {
return ret;
}
-static int alac_encode_frame(AVCodecContext *avctx, uint8_t *frame,
- int buf_size, void *data)
+static int alac_encode_frame(AVCodecContext *avctx, AVPacket *avpkt,
+ const AVFrame *frame, int *got_packet_ptr)
{
AlacEncodeContext *s = avctx->priv_data;
- PutBitContext *pb = &s->pbctx;
- int i, out_bytes, verbatim_flag = 0;
+ int out_bytes, max_frame_size, ret;
+ int16_t **samples = (int16_t **)frame->extended_data;
- if (buf_size < 2 * s->max_coded_frame_size) {
- av_log(avctx, AV_LOG_ERROR, "buffer size is too small\n");
- return -1;
- }
+ s->frame_size = frame->nb_samples;
-verbatim:
- init_put_bits(pb, frame, buf_size);
+ if (frame->nb_samples < DEFAULT_FRAME_SIZE)
+ max_frame_size = get_max_frame_size(s->frame_size, avctx->channels,
+ DEFAULT_SAMPLE_SIZE);
+ else
+ max_frame_size = s->max_coded_frame_size;
- if (s->compression_level == 0 || verbatim_flag) {
- // Verbatim mode
- const int16_t *samples = data;
- write_frame_header(s, 1);
- for (i = 0; i < avctx->frame_size * avctx->channels; i++) {
- put_sbits(pb, 16, *samples++);
- }
- } else {
- init_sample_buffers(s, data);
- write_frame_header(s, 0);
- write_compressed_frame(s);
+ if ((ret = ff_alloc_packet(avpkt, 2 * max_frame_size))) {
+ av_log(avctx, AV_LOG_ERROR, "Error getting output packet\n");
+ return ret;
}
- put_bits(pb, 3, 7);
- flush_put_bits(pb);
- out_bytes = put_bits_count(pb) >> 3;
+ /* use verbatim mode for compression_level 0 */
+ s->verbatim = !s->compression_level;
+
+ out_bytes = write_frame(s, avpkt, samples);
- if (out_bytes > s->max_coded_frame_size) {
+ if (out_bytes > max_frame_size) {
/* frame too large. use verbatim mode */
- if (verbatim_flag || s->compression_level == 0) {
- /* still too large. must be an error. */
- av_log(avctx, AV_LOG_ERROR, "error encoding frame\n");
- return -1;
- }
- verbatim_flag = 1;
- goto verbatim;
+ s->verbatim = 1;
+ out_bytes = write_frame(s, avpkt, samples);
}
- return out_bytes;
+ avpkt->size = out_bytes;
+ *got_packet_ptr = 1;
+ return 0;
}
AVCodec ff_alac_encoder = {
.name = "alac",
.type = AVMEDIA_TYPE_AUDIO,
- .id = CODEC_ID_ALAC,
+ .id = AV_CODEC_ID_ALAC,
.priv_data_size = sizeof(AlacEncodeContext),
.init = alac_encode_init,
- .encode = alac_encode_frame,
+ .encode2 = alac_encode_frame,
.close = alac_encode_close,
- .capabilities = CODEC_CAP_SMALL_LAST_FRAME,
- .sample_fmts = (const enum AVSampleFormat[]){ AV_SAMPLE_FMT_S16,
- AV_SAMPLE_FMT_NONE },
- .long_name = NULL_IF_CONFIG_SMALL("ALAC (Apple Lossless Audio Codec)"),
+ .capabilities = CODEC_CAP_SMALL_LAST_FRAME,
+ .sample_fmts = (const enum AVSampleFormat[]){ AV_SAMPLE_FMT_S16P,
+ AV_SAMPLE_FMT_NONE },
+ .long_name = NULL_IF_CONFIG_SMALL("ALAC (Apple Lossless Audio Codec)"),
};