#define AAC_MAX_CHANNELS 6
+#define ERROR_IF(cond, ...) \
+ if (cond) { \
+ av_log(avctx, AV_LOG_ERROR, __VA_ARGS__); \
+ return AVERROR(EINVAL); \
+ }
+
+float ff_aac_pow34sf_tab[428];
+
static const uint8_t swb_size_1024_96[] = {
4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 8, 8, 8, 8, 8,
12, 12, 12, 12, 12, 16, 16, 24, 28, 36, 44,
{4, TYPE_SCE, TYPE_CPE, TYPE_CPE, TYPE_LFE}, // 6 channels - front center + stereo + back stereo + LFE
};
-static const uint8_t channel_maps[][AAC_MAX_CHANNELS] = {
+/**
+ * Table to remap channels from Libav's default order to AAC order.
+ */
+static const uint8_t aac_chan_maps[AAC_MAX_CHANNELS][AAC_MAX_CHANNELS] = {
{ 0 },
{ 0, 1 },
{ 2, 0, 1 },
init_put_bits(&pb, avctx->extradata, avctx->extradata_size*8);
put_bits(&pb, 5, 2); //object type - AAC-LC
put_bits(&pb, 4, s->samplerate_index); //sample rate index
- put_bits(&pb, 4, avctx->channels);
+ put_bits(&pb, 4, s->channels);
//GASpecificConfig
put_bits(&pb, 1, 0); //frame length - 1024 samples
put_bits(&pb, 1, 0); //does not depend on core coder
flush_put_bits(&pb);
}
-static av_cold int aac_encode_init(AVCodecContext *avctx)
-{
- AACEncContext *s = avctx->priv_data;
- int i;
- const uint8_t *sizes[2];
- uint8_t grouping[AAC_MAX_CHANNELS];
- int lengths[2];
-
- avctx->frame_size = 1024;
+#define WINDOW_FUNC(type) \
+static void apply_ ##type ##_window(DSPContext *dsp, SingleChannelElement *sce, const float *audio)
- for (i = 0; i < 16; i++)
- if (avctx->sample_rate == avpriv_mpeg4audio_sample_rates[i])
- break;
- if (i == 16) {
- av_log(avctx, AV_LOG_ERROR, "Unsupported sample rate %d\n", avctx->sample_rate);
- return -1;
- }
- if (avctx->channels > AAC_MAX_CHANNELS) {
- av_log(avctx, AV_LOG_ERROR, "Unsupported number of channels: %d\n", avctx->channels);
- return -1;
- }
- if (avctx->profile != FF_PROFILE_UNKNOWN && avctx->profile != FF_PROFILE_AAC_LOW) {
- av_log(avctx, AV_LOG_ERROR, "Unsupported profile %d\n", avctx->profile);
- return -1;
- }
- if (1024.0 * avctx->bit_rate / avctx->sample_rate > 6144 * avctx->channels) {
- av_log(avctx, AV_LOG_ERROR, "Too many bits per frame requested\n");
- return -1;
- }
- s->samplerate_index = i;
-
- dsputil_init(&s->dsp, avctx);
- ff_mdct_init(&s->mdct1024, 11, 0, 1.0);
- ff_mdct_init(&s->mdct128, 8, 0, 1.0);
- // window init
- ff_kbd_window_init(ff_aac_kbd_long_1024, 4.0, 1024);
- ff_kbd_window_init(ff_aac_kbd_short_128, 6.0, 128);
- ff_init_ff_sine_windows(10);
- ff_init_ff_sine_windows(7);
-
- s->chan_map = aac_chan_configs[avctx->channels-1];
- s->samples = av_malloc(2 * 1024 * avctx->channels * sizeof(s->samples[0]));
- s->cpe = av_mallocz(sizeof(ChannelElement) * s->chan_map[0]);
- avctx->extradata = av_mallocz(5 + FF_INPUT_BUFFER_PADDING_SIZE);
- avctx->extradata_size = 5;
- put_audio_specific_config(avctx);
+WINDOW_FUNC(only_long)
+{
+ const float *lwindow = sce->ics.use_kb_window[0] ? ff_aac_kbd_long_1024 : ff_sine_1024;
+ const float *pwindow = sce->ics.use_kb_window[1] ? ff_aac_kbd_long_1024 : ff_sine_1024;
+ float *out = sce->ret;
- sizes[0] = swb_size_1024[i];
- sizes[1] = swb_size_128[i];
- lengths[0] = ff_aac_num_swb_1024[i];
- lengths[1] = ff_aac_num_swb_128[i];
- for (i = 0; i < s->chan_map[0]; i++)
- grouping[i] = s->chan_map[i + 1] == TYPE_CPE;
- ff_psy_init(&s->psy, avctx, 2, sizes, lengths, s->chan_map[0], grouping);
- s->psypp = ff_psy_preprocess_init(avctx);
- s->coder = &ff_aac_coders[s->options.aac_coder];
+ dsp->vector_fmul (out, audio, lwindow, 1024);
+ dsp->vector_fmul_reverse(out + 1024, audio + 1024, pwindow, 1024);
+}
- s->lambda = avctx->global_quality ? avctx->global_quality : 120;
+WINDOW_FUNC(long_start)
+{
+ const float *lwindow = sce->ics.use_kb_window[1] ? ff_aac_kbd_long_1024 : ff_sine_1024;
+ const float *swindow = sce->ics.use_kb_window[0] ? ff_aac_kbd_short_128 : ff_sine_128;
+ float *out = sce->ret;
+
+ dsp->vector_fmul(out, audio, lwindow, 1024);
+ memcpy(out + 1024, audio, sizeof(out[0]) * 448);
+ dsp->vector_fmul_reverse(out + 1024 + 448, audio, swindow, 128);
+ memset(out + 1024 + 576, 0, sizeof(out[0]) * 448);
+}
- ff_aac_tableinit();
+WINDOW_FUNC(long_stop)
+{
+ const float *lwindow = sce->ics.use_kb_window[0] ? ff_aac_kbd_long_1024 : ff_sine_1024;
+ const float *swindow = sce->ics.use_kb_window[1] ? ff_aac_kbd_short_128 : ff_sine_128;
+ float *out = sce->ret;
+
+ memset(out, 0, sizeof(out[0]) * 448);
+ dsp->vector_fmul(out + 448, audio + 448, swindow, 128);
+ memcpy(out + 576, audio + 576, sizeof(out[0]) * 448);
+ dsp->vector_fmul_reverse(out + 1024, audio + 1024, lwindow, 1024);
+}
- return 0;
+WINDOW_FUNC(eight_short)
+{
+ const float *swindow = sce->ics.use_kb_window[0] ? ff_aac_kbd_short_128 : ff_sine_128;
+ const float *pwindow = sce->ics.use_kb_window[1] ? ff_aac_kbd_short_128 : ff_sine_128;
+ const float *in = audio + 448;
+ float *out = sce->ret;
+
+ for (int w = 0; w < 8; w++) {
+ dsp->vector_fmul (out, in, w ? pwindow : swindow, 128);
+ out += 128;
+ in += 128;
+ dsp->vector_fmul_reverse(out, in, swindow, 128);
+ out += 128;
+ }
}
-static void apply_window_and_mdct(AVCodecContext *avctx, AACEncContext *s,
- SingleChannelElement *sce, short *audio)
+static void (*const apply_window[4])(DSPContext *dsp, SingleChannelElement *sce, const float *audio) = {
+ [ONLY_LONG_SEQUENCE] = apply_only_long_window,
+ [LONG_START_SEQUENCE] = apply_long_start_window,
+ [EIGHT_SHORT_SEQUENCE] = apply_eight_short_window,
+ [LONG_STOP_SEQUENCE] = apply_long_stop_window
+};
+
+static void apply_window_and_mdct(AACEncContext *s, SingleChannelElement *sce,
+ float *audio)
{
- int i, k;
- const int chans = avctx->channels;
- const float * lwindow = sce->ics.use_kb_window[0] ? ff_aac_kbd_long_1024 : ff_sine_1024;
- const float * swindow = sce->ics.use_kb_window[0] ? ff_aac_kbd_short_128 : ff_sine_128;
- const float * pwindow = sce->ics.use_kb_window[1] ? ff_aac_kbd_short_128 : ff_sine_128;
+ int i;
float *output = sce->ret;
- if (sce->ics.window_sequence[0] != EIGHT_SHORT_SEQUENCE) {
- memcpy(output, sce->saved, sizeof(float)*1024);
- if (sce->ics.window_sequence[0] == LONG_STOP_SEQUENCE) {
- memset(output, 0, sizeof(output[0]) * 448);
- for (i = 448; i < 576; i++)
- output[i] = sce->saved[i] * pwindow[i - 448];
- for (i = 576; i < 704; i++)
- output[i] = sce->saved[i];
- }
- if (sce->ics.window_sequence[0] != LONG_START_SEQUENCE) {
- for (i = 0; i < 1024; i++) {
- output[i+1024] = audio[i * chans] * lwindow[1024 - i - 1];
- sce->saved[i] = audio[i * chans] * lwindow[i];
- }
- } else {
- for (i = 0; i < 448; i++)
- output[i+1024] = audio[i * chans];
- for (; i < 576; i++)
- output[i+1024] = audio[i * chans] * swindow[576 - i - 1];
- memset(output+1024+576, 0, sizeof(output[0]) * 448);
- for (i = 0; i < 1024; i++)
- sce->saved[i] = audio[i * chans];
- }
+ apply_window[sce->ics.window_sequence[0]](&s->dsp, sce, audio);
+
+ if (sce->ics.window_sequence[0] != EIGHT_SHORT_SEQUENCE)
s->mdct1024.mdct_calc(&s->mdct1024, sce->coeffs, output);
- } else {
- for (k = 0; k < 1024; k += 128) {
- for (i = 448 + k; i < 448 + k + 256; i++)
- output[i - 448 - k] = (i < 1024)
- ? sce->saved[i]
- : audio[(i-1024)*chans];
- s->dsp.vector_fmul (output, output, k ? swindow : pwindow, 128);
- s->dsp.vector_fmul_reverse(output+128, output+128, swindow, 128);
- s->mdct128.mdct_calc(&s->mdct128, sce->coeffs + k, output);
- }
- for (i = 0; i < 1024; i++)
- sce->saved[i] = audio[i * chans];
- }
+ else
+ for (i = 0; i < 1024; i += 128)
+ s->mdct128.mdct_calc(&s->mdct128, sce->coeffs + i, output + i*2);
+ memcpy(audio, audio + 1024, sizeof(audio[0]) * 1024);
}
/**
put_bits(&s->pb, 3, TYPE_FIL);
put_bits(&s->pb, 4, FFMIN(namelen, 15));
if (namelen >= 15)
- put_bits(&s->pb, 8, namelen - 16);
+ put_bits(&s->pb, 8, namelen - 14);
put_bits(&s->pb, 4, 0); //extension type - filler
padbits = 8 - (put_bits_count(&s->pb) & 7);
avpriv_align_put_bits(&s->pb);
put_bits(&s->pb, 12 - padbits, 0);
}
+/*
+ * Deinterleave input samples.
+ * Channels are reordered from Libav's default order to AAC order.
+ */
+static void deinterleave_input_samples(AACEncContext *s,
+ const float *samples)
+{
+ int ch, i;
+ const int sinc = s->channels;
+ const uint8_t *channel_map = aac_chan_maps[sinc - 1];
+
+ /* deinterleave and remap input samples */
+ for (ch = 0; ch < sinc; ch++) {
+ const float *sptr = samples + channel_map[ch];
+
+ /* copy last 1024 samples of previous frame to the start of the current frame */
+ memcpy(&s->planar_samples[ch][0], &s->planar_samples[ch][1024], 1024 * sizeof(s->planar_samples[0][0]));
+
+ /* deinterleave */
+ for (i = 1024; i < 1024 * 2; i++) {
+ s->planar_samples[ch][i] = *sptr;
+ sptr += sinc;
+ }
+ }
+}
+
static int aac_encode_frame(AVCodecContext *avctx,
uint8_t *frame, int buf_size, void *data)
{
AACEncContext *s = avctx->priv_data;
- int16_t *samples = s->samples, *samples2, *la;
+ float **samples = s->planar_samples, *samples2, *la, *overlap;
ChannelElement *cpe;
int i, ch, w, g, chans, tag, start_ch;
int chan_el_counter[4];
if (s->last_frame)
return 0;
+
if (data) {
- if (!s->psypp) {
- if (avctx->channels <= 2) {
- memcpy(s->samples + 1024 * avctx->channels, data,
- 1024 * avctx->channels * sizeof(s->samples[0]));
- } else {
- for (i = 0; i < 1024; i++)
- for (ch = 0; ch < avctx->channels; ch++)
- s->samples[(i + 1024) * avctx->channels + ch] =
- ((int16_t*)data)[i * avctx->channels +
- channel_maps[avctx->channels-1][ch]];
- }
- } else {
- start_ch = 0;
- samples2 = s->samples + 1024 * avctx->channels;
- for (i = 0; i < s->chan_map[0]; i++) {
- tag = s->chan_map[i+1];
- chans = tag == TYPE_CPE ? 2 : 1;
- ff_psy_preprocess(s->psypp,
- (uint16_t*)data + channel_maps[avctx->channels-1][start_ch],
- samples2 + start_ch, start_ch, chans);
- start_ch += chans;
- }
- }
+ deinterleave_input_samples(s, data);
+ if (s->psypp)
+ ff_psy_preprocess(s->psypp, s->planar_samples, s->channels);
}
- if (!avctx->frame_number) {
- memcpy(s->samples, s->samples + 1024 * avctx->channels,
- 1024 * avctx->channels * sizeof(s->samples[0]));
+
+ if (!avctx->frame_number)
return 0;
- }
start_ch = 0;
for (i = 0; i < s->chan_map[0]; i++) {
for (ch = 0; ch < chans; ch++) {
IndividualChannelStream *ics = &cpe->ch[ch].ics;
int cur_channel = start_ch + ch;
- samples2 = samples + cur_channel;
- la = samples2 + (448+64) * avctx->channels;
+ overlap = &samples[cur_channel][0];
+ samples2 = overlap + 1024;
+ la = samples2 + (448+64);
if (!data)
la = NULL;
if (tag == TYPE_LFE) {
for (w = 0; w < ics->num_windows; w++)
ics->group_len[w] = wi[ch].grouping[w];
- apply_window_and_mdct(avctx, s, &cpe->ch[ch], samples2);
+ apply_window_and_mdct(s, &cpe->ch[ch], overlap);
}
start_ch += chans;
}
}
frame_bits = put_bits_count(&s->pb);
- if (frame_bits <= 6144 * avctx->channels - 3) {
- s->psy.bitres.bits = frame_bits / avctx->channels;
+ if (frame_bits <= 6144 * s->channels - 3) {
+ s->psy.bitres.bits = frame_bits / s->channels;
break;
}
if (!data)
s->last_frame = 1;
- memcpy(s->samples, s->samples + 1024 * avctx->channels,
- 1024 * avctx->channels * sizeof(s->samples[0]));
+
return put_bits_count(&s->pb)>>3;
}
ff_mdct_end(&s->mdct1024);
ff_mdct_end(&s->mdct128);
ff_psy_end(&s->psy);
- ff_psy_preprocess_end(s->psypp);
- av_freep(&s->samples);
+ if (s->psypp)
+ ff_psy_preprocess_end(s->psypp);
+ av_freep(&s->buffer.samples);
av_freep(&s->cpe);
return 0;
}
+static av_cold int dsp_init(AVCodecContext *avctx, AACEncContext *s)
+{
+ int ret = 0;
+
+ dsputil_init(&s->dsp, avctx);
+
+ // window init
+ ff_kbd_window_init(ff_aac_kbd_long_1024, 4.0, 1024);
+ ff_kbd_window_init(ff_aac_kbd_short_128, 6.0, 128);
+ ff_init_ff_sine_windows(10);
+ ff_init_ff_sine_windows(7);
+
+ if (ret = ff_mdct_init(&s->mdct1024, 11, 0, 32768.0))
+ return ret;
+ if (ret = ff_mdct_init(&s->mdct128, 8, 0, 32768.0))
+ return ret;
+
+ return 0;
+}
+
+static av_cold int alloc_buffers(AVCodecContext *avctx, AACEncContext *s)
+{
+ FF_ALLOCZ_OR_GOTO(avctx, s->buffer.samples, 3 * 1024 * s->channels * sizeof(s->buffer.samples[0]), alloc_fail);
+ FF_ALLOCZ_OR_GOTO(avctx, s->cpe, sizeof(ChannelElement) * s->chan_map[0], alloc_fail);
+ FF_ALLOCZ_OR_GOTO(avctx, avctx->extradata, 5 + FF_INPUT_BUFFER_PADDING_SIZE, alloc_fail);
+
+ for(int ch = 0; ch < s->channels; ch++)
+ s->planar_samples[ch] = s->buffer.samples + 3 * 1024 * ch;
+
+ return 0;
+alloc_fail:
+ return AVERROR(ENOMEM);
+}
+
+static av_cold int aac_encode_init(AVCodecContext *avctx)
+{
+ AACEncContext *s = avctx->priv_data;
+ int i, ret = 0;
+ const uint8_t *sizes[2];
+ uint8_t grouping[AAC_MAX_CHANNELS];
+ int lengths[2];
+
+ avctx->frame_size = 1024;
+
+ for (i = 0; i < 16; i++)
+ if (avctx->sample_rate == avpriv_mpeg4audio_sample_rates[i])
+ break;
+
+ s->channels = avctx->channels;
+
+ ERROR_IF(i == 16,
+ "Unsupported sample rate %d\n", avctx->sample_rate);
+ ERROR_IF(s->channels > AAC_MAX_CHANNELS,
+ "Unsupported number of channels: %d\n", s->channels);
+ ERROR_IF(avctx->profile != FF_PROFILE_UNKNOWN && avctx->profile != FF_PROFILE_AAC_LOW,
+ "Unsupported profile %d\n", avctx->profile);
+ ERROR_IF(1024.0 * avctx->bit_rate / avctx->sample_rate > 6144 * s->channels,
+ "Too many bits per frame requested\n");
+
+ s->samplerate_index = i;
+
+ s->chan_map = aac_chan_configs[s->channels-1];
+
+ if (ret = dsp_init(avctx, s))
+ goto fail;
+
+ if (ret = alloc_buffers(avctx, s))
+ goto fail;
+
+ avctx->extradata_size = 5;
+ put_audio_specific_config(avctx);
+
+ sizes[0] = swb_size_1024[i];
+ sizes[1] = swb_size_128[i];
+ lengths[0] = ff_aac_num_swb_1024[i];
+ lengths[1] = ff_aac_num_swb_128[i];
+ for (i = 0; i < s->chan_map[0]; i++)
+ grouping[i] = s->chan_map[i + 1] == TYPE_CPE;
+ if (ret = ff_psy_init(&s->psy, avctx, 2, sizes, lengths, s->chan_map[0], grouping))
+ goto fail;
+ s->psypp = ff_psy_preprocess_init(avctx);
+ s->coder = &ff_aac_coders[s->options.aac_coder];
+
+ s->lambda = avctx->global_quality ? avctx->global_quality : 120;
+
+ ff_aac_tableinit();
+
+ for (i = 0; i < 428; i++)
+ ff_aac_pow34sf_tab[i] = sqrt(ff_aac_pow2sf_tab[i] * sqrt(ff_aac_pow2sf_tab[i]));
+
+ return 0;
+fail:
+ aac_encode_end(avctx);
+ return ret;
+}
+
#define AACENC_FLAGS AV_OPT_FLAG_ENCODING_PARAM | AV_OPT_FLAG_AUDIO_PARAM
static const AVOption aacenc_options[] = {
{"stereo_mode", "Stereo coding method", offsetof(AACEncContext, options.stereo_mode), AV_OPT_TYPE_INT, {.dbl = 0}, -1, 1, AACENC_FLAGS, "stereo_mode"},
.encode = aac_encode_frame,
.close = aac_encode_end,
.capabilities = CODEC_CAP_SMALL_LAST_FRAME | CODEC_CAP_DELAY | CODEC_CAP_EXPERIMENTAL,
- .sample_fmts = (const enum AVSampleFormat[]){AV_SAMPLE_FMT_S16,AV_SAMPLE_FMT_NONE},
+ .sample_fmts = (const enum AVSampleFormat[]){AV_SAMPLE_FMT_FLT,AV_SAMPLE_FMT_NONE},
.long_name = NULL_IF_CONFIG_SMALL("Advanced Audio Coding"),
.priv_class = &aacenc_class,
};
projects / ffmpeg / blobdiff