generic-library/ffmpeg
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

Merge commit '3d3cf6745e2a5dc9c377244454c3186d75b177fa'

Browse Source 
* commit '3d3cf6745e2a5dc9c377244454c3186d75b177fa':
  aacdec: use float planar sample format for output

Conflicts:
	libavcodec/aacdec.c
	libavcodec/aacsbr.c

Merged-by: Michael Niedermayer <michaelni@gmx.at>
This commit is contained in:
Michael Niedermayer 2012-11-26 15:15:02 +01:00
commit 59b68ee887
4 changed files with 76 additions and 56 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

7
libavcodec/aac.h
View File

@ -236,9 +236,10 @@ typedef struct SingleChannelElement {
uint8_t zeroes[128]; ///< band is not coded (used by encoder) uint8_t zeroes[128]; ///< band is not coded (used by encoder)
DECLARE_ALIGNED(32, float, coeffs)[1024]; ///< coefficients for IMDCT DECLARE_ALIGNED(32, float, coeffs)[1024]; ///< coefficients for IMDCT
DECLARE_ALIGNED(32, float, saved)[1024]; ///< overlap DECLARE_ALIGNED(32, float, saved)[1024]; ///< overlap
DECLARE_ALIGNED(32, float, ret)[2048]; ///< PCM output DECLARE_ALIGNED(32, float, ret_buf)[2048]; ///< PCM output buffer
DECLARE_ALIGNED(16, float, ltp_state)[3072]; ///< time signal for LTP DECLARE_ALIGNED(16, float, ltp_state)[3072]; ///< time signal for LTP
PredictorState predictor_state[MAX_PREDICTORS]; PredictorState predictor_state[MAX_PREDICTORS];
float *ret; ///< PCM output
} SingleChannelElement; } SingleChannelElement;
/** /**
@ -297,10 +298,10 @@ typedef struct AACContext {
/** @} */ /** @} */
/** /**
* @name Members used for output interleaving * @name Members used for output
* @{ * @{
*/ */
float *output_data[MAX_CHANNELS]; ///< Points to each element's 'ret' buffer (PCM output). SingleChannelElement *output_element[MAX_CHANNELS]; ///< Points to each SingleChannelElement
/** @} */ /** @} */

109
libavcodec/aacdec.c
View File

@ -153,10 +153,10 @@ static av_cold int che_configure(AACContext *ac,
av_log(ac->avctx, AV_LOG_ERROR, "Too many channels\n"); av_log(ac->avctx, AV_LOG_ERROR, "Too many channels\n");
return AVERROR_INVALIDDATA; return AVERROR_INVALIDDATA;
} }
ac->output_data[(*channels)++] = ac->che[type][id]->ch[0].ret; ac->output_element[(*channels)++] = &ac->che[type][id]->ch[0];
if (type == TYPE_CPE || if (type == TYPE_CPE ||
(type == TYPE_SCE && ac->oc[1].m4ac.ps == 1)) { (type == TYPE_SCE && ac->oc[1].m4ac.ps == 1)) {
ac->output_data[(*channels)++] = ac->che[type][id]->ch[1].ret; ac->output_element[(*channels)++] = &ac->che[type][id]->ch[1];
} }
} }
} else { } else {
@ -167,6 +167,38 @@ static av_cold int che_configure(AACContext *ac,
return 0; return 0;
} }
static int frame_configure_elements(AVCodecContext *avctx)
{
AACContext *ac = avctx->priv_data;
int type, id, ch, ret;
/* set channel pointers to internal buffers by default */
for (type = 0; type < 4; type++) {
for (id = 0; id < MAX_ELEM_ID; id++) {
ChannelElement *che = ac->che[type][id];
if (che) {
che->ch[0].ret = che->ch[0].ret_buf;
che->ch[1].ret = che->ch[1].ret_buf;
}
}
}
/* get output buffer */
ac->frame.nb_samples = 2048;
if ((ret = avctx->get_buffer(avctx, &ac->frame)) < 0) {
av_log(avctx, AV_LOG_ERROR, "get_buffer() failed\n");
return ret;
}
/* map output channel pointers to AVFrame data */
for (ch = 0; ch < avctx->channels; ch++) {
if (ac->output_element[ch])
ac->output_element[ch]->ret = (float *)ac->frame.extended_data[ch];
}
return 0;
}
struct elem_to_channel { struct elem_to_channel {
uint64_t av_position; uint64_t av_position;
uint8_t syn_ele; uint8_t syn_ele;
@ -382,8 +414,8 @@ static void pop_output_configuration(AACContext *ac) {
* @return Returns error status. 0 - OK, !0 - error * @return Returns error status. 0 - OK, !0 - error
*/ */
static int output_configure(AACContext *ac, static int output_configure(AACContext *ac,
uint8_t layout_map[MAX_ELEM_ID*4][3], int tags, uint8_t layout_map[MAX_ELEM_ID*4][3], int tags,
enum OCStatus oc_type) enum OCStatus oc_type, int get_new_frame)
{ {
AVCodecContext *avctx = ac->avctx; AVCodecContext *avctx = ac->avctx;
int i, channels = 0, ret; int i, channels = 0, ret;
@ -422,6 +454,11 @@ static int output_configure(AACContext *ac,
avctx->channels = ac->oc[1].channels = channels; avctx->channels = ac->oc[1].channels = channels;
ac->oc[1].status = oc_type; ac->oc[1].status = oc_type;
if (get_new_frame) {
if ((ret = frame_configure_elements(ac->avctx)) < 0)
return ret;
}
return 0; return 0;
} }
@ -481,7 +518,7 @@ static ChannelElement *get_che(AACContext *ac, int type, int elem_id)
2) < 0) 2) < 0)
return NULL; return NULL;
if (output_configure(ac, layout_map, layout_map_tags, if (output_configure(ac, layout_map, layout_map_tags,
OC_TRIAL_FRAME) < 0) OC_TRIAL_FRAME, 1) < 0)
return NULL; return NULL;
ac->oc[1].m4ac.chan_config = 2; ac->oc[1].m4ac.chan_config = 2;
@ -499,7 +536,7 @@ static ChannelElement *get_che(AACContext *ac, int type, int elem_id)
1) < 0) 1) < 0)
return NULL; return NULL;
if (output_configure(ac, layout_map, layout_map_tags, if (output_configure(ac, layout_map, layout_map_tags,
OC_TRIAL_FRAME) < 0) OC_TRIAL_FRAME, 1) < 0)
return NULL; return NULL;
ac->oc[1].m4ac.chan_config = 1; ac->oc[1].m4ac.chan_config = 1;
@ -692,7 +729,7 @@ static int decode_ga_specific_config(AACContext *ac, AVCodecContext *avctx,
} else if (m4ac->sbr == 1 && m4ac->ps == -1) } else if (m4ac->sbr == 1 && m4ac->ps == -1)
m4ac->ps = 1; m4ac->ps = 1;
if (ac && (ret = output_configure(ac, layout_map, tags, OC_GLOBAL_HDR))) if (ac && (ret = output_configure(ac, layout_map, tags, OC_GLOBAL_HDR, 0)))
return ret; return ret;
if (extension_flag) { if (extension_flag) {
@ -834,18 +871,11 @@ static void reset_predictor_group(PredictorState *ps, int group_num)
static av_cold int aac_decode_init(AVCodecContext *avctx) static av_cold int aac_decode_init(AVCodecContext *avctx)
{ {
AACContext *ac = avctx->priv_data; AACContext *ac = avctx->priv_data;
float output_scale_factor;
ac->avctx = avctx; ac->avctx = avctx;
ac->oc[1].m4ac.sample_rate = avctx->sample_rate; ac->oc[1].m4ac.sample_rate = avctx->sample_rate;
if (avctx->request_sample_fmt == AV_SAMPLE_FMT_FLT) { avctx->sample_fmt = AV_SAMPLE_FMT_FLTP;
avctx->sample_fmt = AV_SAMPLE_FMT_FLT;
output_scale_factor = 1.0 / 32768.0;
} else {
avctx->sample_fmt = AV_SAMPLE_FMT_S16;
output_scale_factor = 1.0;
}
if (avctx->extradata_size > 0) { if (avctx->extradata_size > 0) {
if (decode_audio_specific_config(ac, ac->avctx, &ac->oc[1].m4ac, if (decode_audio_specific_config(ac, ac->avctx, &ac->oc[1].m4ac,
@ -876,7 +906,7 @@ static av_cold int aac_decode_init(AVCodecContext *avctx)
&layout_map_tags, ac->oc[1].m4ac.chan_config); &layout_map_tags, ac->oc[1].m4ac.chan_config);
if (!ret) if (!ret)
output_configure(ac, layout_map, layout_map_tags, output_configure(ac, layout_map, layout_map_tags,
OC_GLOBAL_HDR); OC_GLOBAL_HDR, 0);
else if (avctx->err_recognition & AV_EF_EXPLODE) else if (avctx->err_recognition & AV_EF_EXPLODE)
return AVERROR_INVALIDDATA; return AVERROR_INVALIDDATA;
} }
@ -909,9 +939,9 @@ static av_cold int aac_decode_init(AVCodecContext *avctx)
ff_aac_scalefactor_code, sizeof(ff_aac_scalefactor_code[0]), sizeof(ff_aac_scalefactor_code[0]), ff_aac_scalefactor_code, sizeof(ff_aac_scalefactor_code[0]), sizeof(ff_aac_scalefactor_code[0]),
352); 352);
ff_mdct_init(&ac->mdct, 11, 1, output_scale_factor/1024.0); ff_mdct_init(&ac->mdct, 11, 1, 1.0 / (32768.0 * 1024.0));
ff_mdct_init(&ac->mdct_small, 8, 1, output_scale_factor/128.0); ff_mdct_init(&ac->mdct_small, 8, 1, 1.0 / (32768.0 * 128.0));
ff_mdct_init(&ac->mdct_ltp, 11, 0, -2.0/output_scale_factor); ff_mdct_init(&ac->mdct_ltp, 11, 0, -2.0 * 32768.0);
// window initialization // window initialization
ff_kbd_window_init(ff_aac_kbd_long_1024, 4.0, 1024); 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_kbd_window_init(ff_aac_kbd_short_128, 6.0, 128);
@ -2001,7 +2031,7 @@ static int decode_extension_payload(AACContext *ac, GetBitContext *gb, int cnt,
ac->oc[1].m4ac.sbr = 1; ac->oc[1].m4ac.sbr = 1;
ac->oc[1].m4ac.ps = 1; ac->oc[1].m4ac.ps = 1;
output_configure(ac, ac->oc[1].layout_map, ac->oc[1].layout_map_tags, output_configure(ac, ac->oc[1].layout_map, ac->oc[1].layout_map_tags,
ac->oc[1].status); ac->oc[1].status, 1);
} else { } else {
ac->oc[1].m4ac.sbr = 1; ac->oc[1].m4ac.sbr = 1;
} }
@ -2395,7 +2425,7 @@ static int parse_adts_frame_header(AACContext *ac, GetBitContext *gb)
&layout_map_tags, hdr_info.chan_config)) &layout_map_tags, hdr_info.chan_config))
return -7; return -7;
if (output_configure(ac, layout_map, layout_map_tags, if (output_configure(ac, layout_map, layout_map_tags,
FFMAX(ac->oc[1].status, OC_TRIAL_FRAME))) FFMAX(ac->oc[1].status, OC_TRIAL_FRAME), 0))
return -7; return -7;
} else { } else {
ac->oc[1].m4ac.chan_config = 0; ac->oc[1].m4ac.chan_config = 0;
@ -2404,6 +2434,7 @@ static int parse_adts_frame_header(AACContext *ac, GetBitContext *gb)
* WITHOUT specifying PCE. * WITHOUT specifying PCE.
* thus, set dual mono as default. * thus, set dual mono as default.
*/ */
#if 0
if (ac->enable_jp_dmono && ac->oc[0].status == OC_NONE) { if (ac->enable_jp_dmono && ac->oc[0].status == OC_NONE) {
layout_map_tags = 2; layout_map_tags = 2;
layout_map[0][0] = layout_map[1][0] = TYPE_SCE; layout_map[0][0] = layout_map[1][0] = TYPE_SCE;
@ -2414,6 +2445,7 @@ static int parse_adts_frame_header(AACContext *ac, GetBitContext *gb)
OC_TRIAL_FRAME)) OC_TRIAL_FRAME))
return -7; return -7;
} }
#endif
} }
ac->oc[1].m4ac.sample_rate = hdr_info.sample_rate; ac->oc[1].m4ac.sample_rate = hdr_info.sample_rate;
ac->oc[1].m4ac.sampling_index = hdr_info.sampling_index; ac->oc[1].m4ac.sampling_index = hdr_info.sampling_index;
@ -2454,6 +2486,11 @@ static int aac_decode_frame_int(AVCodecContext *avctx, void *data,
} }
} }
if (frame_configure_elements(avctx) < 0) {
err = -1;
goto fail;
}
ac->tags_mapped = 0; ac->tags_mapped = 0;
// parse // parse
while ((elem_type = get_bits(gb, 3)) != TYPE_END) { while ((elem_type = get_bits(gb, 3)) != TYPE_END) {
@ -2509,7 +2546,7 @@ static int aac_decode_frame_int(AVCodecContext *avctx, void *data,
"Not evaluating a further program_config_element as this construct is dubious at best.\n"); "Not evaluating a further program_config_element as this construct is dubious at best.\n");
pop_output_configuration(ac); pop_output_configuration(ac);
} else { } else {
err = output_configure(ac, layout_map, tags, OC_TRIAL_PCE); err = output_configure(ac, layout_map, tags, OC_TRIAL_PCE, 1);
if (!err) if (!err)
ac->oc[1].m4ac.chan_config = 0; ac->oc[1].m4ac.chan_config = 0;
pce_found = 1; pce_found = 1;
@ -2552,7 +2589,7 @@ static int aac_decode_frame_int(AVCodecContext *avctx, void *data,
multiplier = (ac->oc[1].m4ac.sbr == 1) ? ac->oc[1].m4ac.ext_sample_rate > ac->oc[1].m4ac.sample_rate : 0; multiplier = (ac->oc[1].m4ac.sbr == 1) ? ac->oc[1].m4ac.ext_sample_rate > ac->oc[1].m4ac.sample_rate : 0;
samples <<= multiplier; samples <<= multiplier;
#if 0
/* for dual-mono audio (SCE + SCE) */ /* for dual-mono audio (SCE + SCE) */
is_dmono = ac->enable_jp_dmono && sce_count == 2 && is_dmono = ac->enable_jp_dmono && sce_count == 2 &&
ac->oc[1].channel_layout == (AV_CH_FRONT_LEFT | AV_CH_FRONT_RIGHT); ac->oc[1].channel_layout == (AV_CH_FRONT_LEFT | AV_CH_FRONT_RIGHT);
@ -2566,36 +2603,20 @@ static int aac_decode_frame_int(AVCodecContext *avctx, void *data,
ac->output_data[0] = ac->output_data[1]; ac->output_data[0] = ac->output_data[1];
} }
} }
#endif
if (samples) { if (samples) {
/* get output buffer */
ac->frame.nb_samples = samples; ac->frame.nb_samples = samples;
if ((err = avctx->get_buffer(avctx, &ac->frame)) < 0) {
av_log(avctx, AV_LOG_ERROR, "get_buffer() failed\n");
err = -1;
goto fail;
}
if (avctx->sample_fmt == AV_SAMPLE_FMT_FLT)
ac->fmt_conv.float_interleave((float *)ac->frame.data[0],
(const float **)ac->output_data,
samples, avctx->channels);
else
ac->fmt_conv.float_to_int16_interleave((int16_t *)ac->frame.data[0],
(const float **)ac->output_data,
samples, avctx->channels);
*(AVFrame *)data = ac->frame; *(AVFrame *)data = ac->frame;
} }
*got_frame_ptr = !!samples; *got_frame_ptr = !!samples;
#if 0
if (is_dmono) { if (is_dmono) {
if (ac->dmono_mode == 0) if (ac->dmono_mode == 0)
ac->output_data[1] = tmp; ac->output_data[1] = tmp;
else if (ac->dmono_mode == 1) else if (ac->dmono_mode == 1)
ac->output_data[0] = tmp; ac->output_data[0] = tmp;
} }
#endif
if (ac->oc[1].status && audio_found) { if (ac->oc[1].status && audio_found) {
avctx->sample_rate = ac->oc[1].m4ac.sample_rate << multiplier; avctx->sample_rate = ac->oc[1].m4ac.sample_rate << multiplier;
avctx->frame_size = samples; avctx->frame_size = samples;
@ -2970,7 +2991,7 @@ AVCodec ff_aac_decoder = {
.decode = aac_decode_frame, .decode = aac_decode_frame,
.long_name = NULL_IF_CONFIG_SMALL("AAC (Advanced Audio Coding)"), .long_name = NULL_IF_CONFIG_SMALL("AAC (Advanced Audio Coding)"),
.sample_fmts = (const enum AVSampleFormat[]) { .sample_fmts = (const enum AVSampleFormat[]) {
AV_SAMPLE_FMT_FLT, AV_SAMPLE_FMT_S16, AV_SAMPLE_FMT_NONE AV_SAMPLE_FMT_FLTP, AV_SAMPLE_FMT_NONE
}, },
.capabilities = CODEC_CAP_CHANNEL_CONF | CODEC_CAP_DR1, .capabilities = CODEC_CAP_CHANNEL_CONF | CODEC_CAP_DR1,
.channel_layouts = aac_channel_layout, .channel_layouts = aac_channel_layout,
@ -2992,7 +3013,7 @@ AVCodec ff_aac_latm_decoder = {
.decode = latm_decode_frame, .decode = latm_decode_frame,
.long_name = NULL_IF_CONFIG_SMALL("AAC LATM (Advanced Audio Coding LATM syntax)"), .long_name = NULL_IF_CONFIG_SMALL("AAC LATM (Advanced Audio Coding LATM syntax)"),
.sample_fmts = (const enum AVSampleFormat[]) { .sample_fmts = (const enum AVSampleFormat[]) {
AV_SAMPLE_FMT_FLT, AV_SAMPLE_FMT_S16, AV_SAMPLE_FMT_NONE AV_SAMPLE_FMT_FLTP, AV_SAMPLE_FMT_NONE
}, },
.capabilities = CODEC_CAP_CHANNEL_CONF | CODEC_CAP_DR1, .capabilities = CODEC_CAP_CHANNEL_CONF | CODEC_CAP_DR1,
.channel_layouts = aac_channel_layout, .channel_layouts = aac_channel_layout,

10
libavcodec/aacenc.c
View File

@ -191,7 +191,7 @@ WINDOW_FUNC(only_long)
{ {
const float *lwindow = sce->ics.use_kb_window[0] ? ff_aac_kbd_long_1024 : ff_sine_1024; 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; const float *pwindow = sce->ics.use_kb_window[1] ? ff_aac_kbd_long_1024 : ff_sine_1024;
float *out = sce->ret; float *out = sce->ret_buf;
fdsp->vector_fmul (out, audio, lwindow, 1024); fdsp->vector_fmul (out, audio, lwindow, 1024);
dsp->vector_fmul_reverse(out + 1024, audio + 1024, pwindow, 1024); dsp->vector_fmul_reverse(out + 1024, audio + 1024, pwindow, 1024);
@ -201,7 +201,7 @@ WINDOW_FUNC(long_start)
{ {
const float *lwindow = sce->ics.use_kb_window[1] ? ff_aac_kbd_long_1024 : ff_sine_1024; 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; const float *swindow = sce->ics.use_kb_window[0] ? ff_aac_kbd_short_128 : ff_sine_128;
float *out = sce->ret; float *out = sce->ret_buf;
fdsp->vector_fmul(out, audio, lwindow, 1024); fdsp->vector_fmul(out, audio, lwindow, 1024);
memcpy(out + 1024, audio + 1024, sizeof(out[0]) * 448); memcpy(out + 1024, audio + 1024, sizeof(out[0]) * 448);
@ -213,7 +213,7 @@ WINDOW_FUNC(long_stop)
{ {
const float *lwindow = sce->ics.use_kb_window[0] ? ff_aac_kbd_long_1024 : ff_sine_1024; 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; const float *swindow = sce->ics.use_kb_window[1] ? ff_aac_kbd_short_128 : ff_sine_128;
float *out = sce->ret; float *out = sce->ret_buf;
memset(out, 0, sizeof(out[0]) * 448); memset(out, 0, sizeof(out[0]) * 448);
fdsp->vector_fmul(out + 448, audio + 448, swindow, 128); fdsp->vector_fmul(out + 448, audio + 448, swindow, 128);
@ -226,7 +226,7 @@ WINDOW_FUNC(eight_short)
const float *swindow = sce->ics.use_kb_window[0] ? ff_aac_kbd_short_128 : ff_sine_128; 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 *pwindow = sce->ics.use_kb_window[1] ? ff_aac_kbd_short_128 : ff_sine_128;
const float *in = audio + 448; const float *in = audio + 448;
float *out = sce->ret; float *out = sce->ret_buf;
int w; int w;
for (w = 0; w < 8; w++) { for (w = 0; w < 8; w++) {
@ -251,7 +251,7 @@ static void apply_window_and_mdct(AACEncContext *s, SingleChannelElement *sce,
float *audio) float *audio)
{ {
int i; int i;
float *output = sce->ret; float *output = sce->ret_buf;
apply_window[sce->ics.window_sequence[0]](&s->dsp, &s->fdsp, sce, audio); apply_window[sce->ics.window_sequence[0]](&s->dsp, &s->fdsp, sce, audio);

6
libavcodec/aacsbr.c
View File

@ -142,7 +142,6 @@ static void sbr_turnoff(SpectralBandReplication *sbr) {
av_cold void ff_aac_sbr_ctx_init(AACContext *ac, SpectralBandReplication *sbr) av_cold void ff_aac_sbr_ctx_init(AACContext *ac, SpectralBandReplication *sbr)
{ {
float mdct_scale;
if(sbr->mdct.mdct_bits) if(sbr->mdct.mdct_bits)
return; return;
sbr->kx[0] = sbr->kx[1]; sbr->kx[0] = sbr->kx[1];
@ -152,9 +151,8 @@ av_cold void ff_aac_sbr_ctx_init(AACContext *ac, SpectralBandReplication *sbr)
/* SBR requires samples to be scaled to +/-32768.0 to work correctly. /* SBR requires samples to be scaled to +/-32768.0 to work correctly.
* mdct scale factors are adjusted to scale up from +/-1.0 at analysis * mdct scale factors are adjusted to scale up from +/-1.0 at analysis
* and scale back down at synthesis. */ * and scale back down at synthesis. */
mdct_scale = ac->avctx->sample_fmt == AV_SAMPLE_FMT_FLT ? 32768.0f : 1.0f; ff_mdct_init(&sbr->mdct, 7, 1, 1.0 / (64 * 32768.0));
ff_mdct_init(&sbr->mdct, 7, 1, 1.0 / (64 * mdct_scale)); ff_mdct_init(&sbr->mdct_ana, 7, 1, -2.0 * 32768.0);
ff_mdct_init(&sbr->mdct_ana, 7, 1, -2.0 * mdct_scale);
ff_ps_ctx_init(&sbr->ps); ff_ps_ctx_init(&sbr->ps);
ff_sbrdsp_init(&sbr->dsp); ff_sbrdsp_init(&sbr->dsp);
} }