---
I'm not 100% sure about 1 change. In output_configure() I moved the call
to sniff_channel_order() to above where the layout map and tags are
saved to ac->oc[1]. Otherwise the final layout map is never saved and
cannot be used to map the AVFrame channel pointers. I don't really know
the full consequences of this change though.
libavcodec/aac.h | 11 +---
libavcodec/aacdec.c | 140 ++++++++++++++++++++++++++++++++++----------------
libavcodec/aacsbr.c | 6 +--
3 files changed, 100 insertions(+), 57 deletions(-)
diff --git a/libavcodec/aac.h b/libavcodec/aac.h
index 32baf9d..18f674f 100644
--- a/libavcodec/aac.h
+++ b/libavcodec/aac.h
@@ -237,9 +237,10 @@ typedef struct SingleChannelElement {
uint8_t zeroes[128]; ///< band is not coded
(used by encoder)
DECLARE_ALIGNED(32, float, coeffs)[1024]; ///< coefficients for IMDCT
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
PredictorState predictor_state[MAX_PREDICTORS];
+ float *ret; ///< PCM output
} SingleChannelElement;
/**
@@ -263,6 +264,7 @@ typedef struct ChannelElement {
typedef struct AACContext {
AVCodecContext *avctx;
AVFrame frame;
+ int got_frame_buffer;
int is_saved; ///< Set if elements have stored overlap
from previous frame.
DynamicRangeControl che_drc;
@@ -297,13 +299,6 @@ typedef struct AACContext {
int random_state;
/** @} */
- /**
- * @name Members used for output interleaving
- * @{
- */
- float *output_data[MAX_CHANNELS]; ///< Points to each
element's 'ret' buffer (PCM output).
- /** @} */
-
DECLARE_ALIGNED(32, float, temp)[128];
OutputConfiguration oc[2];
diff --git a/libavcodec/aacdec.c b/libavcodec/aacdec.c
index 4a6d390..9d3c3b1 100644
--- a/libavcodec/aacdec.c
+++ b/libavcodec/aacdec.c
@@ -79,6 +79,7 @@
Parametric Stereo.
*/
+#include "libavutil/audioconvert.h"
#include "libavutil/float_dsp.h"
#include "avcodec.h"
#include "internal.h"
@@ -149,10 +150,10 @@ static av_cold int che_configure(AACContext *ac,
ff_aac_sbr_ctx_init(ac, &ac->che[type][id]->sbr);
}
if (type != TYPE_CCE) {
- ac->output_data[(*channels)++] = ac->che[type][id]->ch[0].ret;
+ (*channels)++;
if (type == TYPE_CPE ||
(type == TYPE_SCE && ac->oc[1].m4ac.ps == 1)) {
- ac->output_data[(*channels)++] = ac->che[type][id]->ch[1].ret;
+ (*channels)++;
}
}
} else {
@@ -163,6 +164,64 @@ static av_cold int che_configure(AACContext *ac,
return 0;
}
+static int frame_configure_elements(AVCodecContext *avctx, int is_init)
+{
+ AACContext *ac = avctx->priv_data;
+ ChannelElement *che;
+ int ch, i, ret;
+ int type, id, pos;
+
+ /* set channel pointers to internal buffers by default */
+ for (type = 0; type < 4; type++) {
+ for (id = 0; id < MAX_ELEM_ID; id++) {
+ 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;
+ }
+ }
+ }
+
+ /* if this function was called during initialization, we do not need to
+ call get_buffer() or map output channel pointers */
+ if (is_init)
+ return 0;
+
+ /* get output buffer */
+ if (!ac->got_frame_buffer ||
+ ac->oc[1].channel_layout != ac->frame.channel_layout) {
+ 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;
+ }
+ ac->got_frame_buffer = 1;
+ }
+
+ /* map output channel pointers to AVFrame data */
+ ch = 0;
+ for (i = 0; i < ac->oc[1].layout_map_tags; i++) {
+ type = ac->oc[1].layout_map[i][0];
+ id = ac->oc[1].layout_map[i][1];
+ pos = ac->oc[1].layout_map[i][2];
+
+ che = ac->che[type][id];
+ if (che && pos && type != TYPE_CCE) {
+ if (ch >= avctx->channels)
+ return AVERROR(EINVAL);
+ che->ch[0].ret = (float *)ac->frame.extended_data[ch++];
+ if (type == TYPE_CPE ||
+ (type == TYPE_SCE && ac->oc[1].m4ac.ps == 1)) {
+ if (ch >= avctx->channels)
+ return AVERROR(EINVAL);
+ che->ch[1].ret = (float *)ac->frame.extended_data[ch++];
+ }
+ }
+ }
+
+ return 0;
+}
+
struct elem_to_channel {
uint64_t av_position;
uint8_t syn_ele;
@@ -378,22 +437,24 @@ static void pop_output_configuration(AACContext *ac) {
* @return Returns error status. 0 - OK, !0 - error
*/
static int output_configure(AACContext *ac,
- uint8_t layout_map[MAX_ELEM_ID*4][3], int
tags,
- int channel_config, enum OCStatus oc_type)
+ uint8_t layout_map[MAX_ELEM_ID*4][3], int tags,
+ int channel_config, enum OCStatus oc_type,
+ int is_init)
{
AVCodecContext *avctx = ac->avctx;
int i, channels = 0, ret;
uint64_t layout = 0;
+ // Try to sniff a reasonable channel order, otherwise output the
+ // channels in the order the PCE declared them.
+ if (avctx->request_channel_layout != AV_CH_LAYOUT_NATIVE)
+ layout = sniff_channel_order(layout_map, tags);
+
if (ac->oc[1].layout_map != layout_map) {
memcpy(ac->oc[1].layout_map, layout_map, tags * sizeof(layout_map[0]));
ac->oc[1].layout_map_tags = tags;
}
- // Try to sniff a reasonable channel order, otherwise output the
- // channels in the order the PCE declared them.
- if (avctx->request_channel_layout != AV_CH_LAYOUT_NATIVE)
- layout = sniff_channel_order(layout_map, tags);
for (i = 0; i < tags; i++) {
int type = layout_map[i][0];
int id = layout_map[i][1];
@@ -417,6 +478,9 @@ static int output_configure(AACContext *ac,
avctx->channels = ac->oc[1].channels = channels;
ac->oc[1].status = oc_type;
+ if ((ret = frame_configure_elements(ac->avctx, is_init)) < 0)
+ return ret;
+
return 0;
}
@@ -457,7 +521,7 @@ static ChannelElement *get_che(AACContext *ac, int type,
int elem_id)
2) < 0)
return NULL;
if (output_configure(ac, layout_map, layout_map_tags,
- 2, OC_TRIAL_FRAME) < 0)
+ 2, OC_TRIAL_FRAME, 0) < 0)
return NULL;
ac->oc[1].m4ac.chan_config = 2;
@@ -473,7 +537,7 @@ static ChannelElement *get_che(AACContext *ac, int type,
int elem_id)
1) < 0)
return NULL;
if (output_configure(ac, layout_map, layout_map_tags,
- 1, OC_TRIAL_FRAME) < 0)
+ 1, OC_TRIAL_FRAME, 0) < 0)
return NULL;
ac->oc[1].m4ac.chan_config = 1;
@@ -661,7 +725,7 @@ static int decode_ga_specific_config(AACContext *ac,
AVCodecContext *avctx,
m4ac->ps = 1;
if (ac && (ret = output_configure(ac, layout_map, tags,
- channel_config, OC_GLOBAL_HDR)))
+ channel_config, OC_GLOBAL_HDR, 1)))
return ret;
if (extension_flag) {
@@ -803,11 +867,12 @@ static void reset_predictor_group(PredictorState *ps, int
group_num)
static av_cold int aac_decode_init(AVCodecContext *avctx)
{
AACContext *ac = avctx->priv_data;
- float output_scale_factor;
ac->avctx = avctx;
ac->oc[1].m4ac.sample_rate = avctx->sample_rate;
+ avctx->sample_fmt = AV_SAMPLE_FMT_FLTP;
+
if (avctx->extradata_size > 0) {
if (decode_audio_specific_config(ac, ac->avctx, &ac->oc[1].m4ac,
avctx->extradata,
@@ -837,20 +902,12 @@ static av_cold int aac_decode_init(AVCodecContext *avctx)
&layout_map_tags, ac->oc[1].m4ac.chan_config);
if (!ret)
output_configure(ac, layout_map, layout_map_tags,
- ac->oc[1].m4ac.chan_config, OC_GLOBAL_HDR);
+ ac->oc[1].m4ac.chan_config, OC_GLOBAL_HDR, 1);
else if (avctx->err_recognition & AV_EF_EXPLODE)
return AVERROR_INVALIDDATA;
}
}
- if (avctx->request_sample_fmt == AV_SAMPLE_FMT_FLT) {
- 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;
- }
-
AAC_INIT_VLC_STATIC( 0, 304);
AAC_INIT_VLC_STATIC( 1, 270);
AAC_INIT_VLC_STATIC( 2, 550);
@@ -878,9 +935,9 @@ 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]),
352);
- ff_mdct_init(&ac->mdct, 11, 1, output_scale_factor/1024.0);
- ff_mdct_init(&ac->mdct_small, 8, 1, output_scale_factor/128.0);
- ff_mdct_init(&ac->mdct_ltp, 11, 0, -2.0/output_scale_factor);
+ ff_mdct_init(&ac->mdct, 11, 1, 1.0 / (32768.0 * 1024.0));
+ ff_mdct_init(&ac->mdct_small, 8, 1, 1.0 / (32768.0 * 128.0));
+ ff_mdct_init(&ac->mdct_ltp, 11, 0, -2.0 * 32768.0);
// window initialization
ff_kbd_window_init(ff_aac_kbd_long_1024, 4.0, 1024);
ff_kbd_window_init(ff_aac_kbd_short_128, 6.0, 128);
@@ -1943,7 +2000,7 @@ static int decode_extension_payload(AACContext *ac,
GetBitContext *gb, int cnt,
ac->oc[1].m4ac.sbr = 1;
ac->oc[1].m4ac.ps = 1;
output_configure(ac, ac->oc[1].layout_map,
ac->oc[1].layout_map_tags,
- ac->oc[1].m4ac.chan_config, ac->oc[1].status);
+ ac->oc[1].m4ac.chan_config, ac->oc[1].status, 0);
} else {
ac->oc[1].m4ac.sbr = 1;
}
@@ -2334,7 +2391,7 @@ static int parse_adts_frame_header(AACContext *ac,
GetBitContext *gb)
return -7;
if (output_configure(ac, layout_map, layout_map_tags,
hdr_info.chan_config,
- FFMAX(ac->oc[1].status, OC_TRIAL_FRAME)))
+ FFMAX(ac->oc[1].status, OC_TRIAL_FRAME), 0))
return -7;
} else {
ac->oc[1].m4ac.chan_config = 0;
@@ -2376,6 +2433,11 @@ static int aac_decode_frame_int(AVCodecContext *avctx,
void *data,
}
}
+ if (frame_configure_elements(avctx, 0) < 0) {
+ err = -1;
+ goto fail;
+ }
+
ac->tags_mapped = 0;
// parse
while ((elem_type = get_bits(gb, 3)) != TYPE_END) {
@@ -2430,7 +2492,7 @@ 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");
pop_output_configuration(ac);
} else {
- err = output_configure(ac, layout_map, tags, 0, OC_TRIAL_PCE);
+ err = output_configure(ac, layout_map, tags, 0, OC_TRIAL_PCE,
0);
pce_found = 1;
}
break;
@@ -2473,23 +2535,7 @@ static int aac_decode_frame_int(AVCodecContext *avctx,
void *data,
samples <<= multiplier;
if (samples) {
- /* get output buffer */
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;
}
*got_frame_ptr = !!samples;
@@ -2521,6 +2567,8 @@ static int aac_decode_frame(AVCodecContext *avctx, void
*data,
AV_PKT_DATA_NEW_EXTRADATA,
&new_extradata_size);
+ ac->got_frame_buffer = 0;
+
if (new_extradata) {
av_free(avctx->extradata);
avctx->extradata = av_mallocz(new_extradata_size +
@@ -2783,6 +2831,8 @@ static int latm_decode_frame(AVCodecContext *avctx, void
*out,
int muxlength, err;
GetBitContext gb;
+ latmctx->aac_ctx.got_frame_buffer = 0;
+
init_get_bits(&gb, avpkt->data, avpkt->size * 8);
// check for LOAS sync word
@@ -2848,7 +2898,7 @@ AVCodec ff_aac_decoder = {
.decode = aac_decode_frame,
.long_name = NULL_IF_CONFIG_SMALL("AAC (Advanced Audio Coding)"),
.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,
.channel_layouts = aac_channel_layout,
@@ -2869,7 +2919,7 @@ AVCodec ff_aac_latm_decoder = {
.decode = latm_decode_frame,
.long_name = NULL_IF_CONFIG_SMALL("AAC LATM (Advanced Audio Coding
LATM syntax)"),
.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,
.channel_layouts = aac_channel_layout,
diff --git a/libavcodec/aacsbr.c b/libavcodec/aacsbr.c
index 5eca115..c1133d8 100644
--- a/libavcodec/aacsbr.c
+++ b/libavcodec/aacsbr.c
@@ -140,7 +140,6 @@ static void sbr_turnoff(SpectralBandReplication *sbr) {
av_cold void ff_aac_sbr_ctx_init(AACContext *ac, SpectralBandReplication *sbr)
{
- float mdct_scale;
sbr->kx[0] = sbr->kx[1];
sbr_turnoff(sbr);
sbr->data[0].synthesis_filterbank_samples_offset = SBR_SYNTHESIS_BUF_SIZE
- (1280 - 128);
@@ -148,9 +147,8 @@ av_cold void ff_aac_sbr_ctx_init(AACContext *ac,
SpectralBandReplication *sbr)
/* SBR requires samples to be scaled to +/-32768.0 to work correctly.
* mdct scale factors are adjusted to scale up from +/-1.0 at analysis
* 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 * mdct_scale));
- ff_mdct_init(&sbr->mdct_ana, 7, 1, -2.0 * mdct_scale);
+ ff_mdct_init(&sbr->mdct, 7, 1, 1.0 / (64 * 32768.0));
+ ff_mdct_init(&sbr->mdct_ana, 7, 1, -2.0 * 32768.0);
ff_ps_ctx_init(&sbr->ps);
ff_sbrdsp_init(&sbr->dsp);
}
--
1.7.1
_______________________________________________
libav-devel mailing list
libav-devel@libav.org
https://lists.libav.org/mailman/listinfo/libav-devel
