int opus_decode_native(OpusDecoder *st, const unsigned char *data, opus_int32 len, opus_val16 *pcm, int frame_size, int decode_fec, int self_delimited, opus_int32 *packet_offset, int soft_clip) { int i, nb_samples; int count, offset; unsigned char toc; int packet_frame_size, packet_bandwidth, packet_mode, packet_stream_channels; /* 48 x 2.5 ms = 120 ms */ opus_int16 size[48]; if (decode_fec<0 || decode_fec>1) return OPUS_BAD_ARG; /* For FEC/PLC, frame_size has to be to have a multiple of 2.5 ms */ if ((decode_fec || len==0 || data==NULL) && frame_size%(st->Fs/400)!=0) return OPUS_BAD_ARG; if (len==0 || data==NULL) { int pcm_count=0; do { int ret; ret = opus_decode_frame(st, NULL, 0, pcm+pcm_count*st->channels, frame_size-pcm_count, 0); if (ret<0) return ret; pcm_count += ret; } while (pcm_count < frame_size); celt_assert(pcm_count == frame_size); if (OPUS_CHECK_ARRAY(pcm, pcm_count*st->channels)) OPUS_PRINT_INT(pcm_count); st->last_packet_duration = pcm_count; return pcm_count; } else if (len<0) return OPUS_BAD_ARG; packet_mode = opus_packet_get_mode(data); packet_bandwidth = opus_packet_get_bandwidth(data); packet_frame_size = opus_packet_get_samples_per_frame(data, st->Fs); packet_stream_channels = opus_packet_get_nb_channels(data); count = opus_packet_parse_impl(data, len, self_delimited, &toc, NULL, size, &offset, packet_offset); if (count<0) return count; data += offset; if (decode_fec) { int duration_copy; int ret; /* If no FEC can be present, run the PLC (recursive call) */ if (frame_size < packet_frame_size || packet_mode == MODE_CELT_ONLY || st->mode == MODE_CELT_ONLY) return opus_decode_native(st, NULL, 0, pcm, frame_size, 0, 0, NULL, soft_clip); /* Otherwise, run the PLC on everything except the size for which we might have FEC */ duration_copy = st->last_packet_duration; if (frame_size-packet_frame_size!=0) { ret = opus_decode_native(st, NULL, 0, pcm, frame_size-packet_frame_size, 0, 0, NULL, soft_clip); if (ret<0) { st->last_packet_duration = duration_copy; return ret; } celt_assert(ret==frame_size-packet_frame_size); } /* Complete with FEC */ st->mode = packet_mode; st->bandwidth = packet_bandwidth; st->frame_size = packet_frame_size; st->stream_channels = packet_stream_channels; ret = opus_decode_frame(st, data, size[0], pcm+st->channels*(frame_size-packet_frame_size), packet_frame_size, 1); if (ret<0) return ret; else { if (OPUS_CHECK_ARRAY(pcm, frame_size*st->channels)) OPUS_PRINT_INT(frame_size); st->last_packet_duration = frame_size; return frame_size; } } if (count*packet_frame_size > frame_size) return OPUS_BUFFER_TOO_SMALL; /* Update the state as the last step to avoid updating it on an invalid packet */ st->mode = packet_mode; st->bandwidth = packet_bandwidth; st->frame_size = packet_frame_size; st->stream_channels = packet_stream_channels; nb_samples=0; for (i=0; i<count; i++) { int ret; ret = opus_decode_frame(st, data, size[i], pcm+nb_samples*st->channels, frame_size-nb_samples, 0); if (ret<0) return ret; celt_assert(ret==packet_frame_size); data += size[i]; nb_samples += ret; } st->last_packet_duration = nb_samples; if (OPUS_CHECK_ARRAY(pcm, nb_samples*st->channels)) OPUS_PRINT_INT(nb_samples); #ifndef FIXED_POINT if (soft_clip) opus_pcm_soft_clip(pcm, nb_samples, st->channels, st->softclip_mem); else st->softclip_mem[0]=st->softclip_mem[1]=0; #endif return nb_samples; }
static int opus_decode_frame(OpusDecoder *st, const unsigned char *data, opus_int32 len, opus_val16 *pcm, int frame_size, int decode_fec) { void *silk_dec; CELTDecoder *celt_dec; int i, silk_ret=0, celt_ret=0; ec_dec dec; opus_int32 silk_frame_size; int pcm_silk_size; VARDECL(opus_int16, pcm_silk); int pcm_transition_silk_size; VARDECL(opus_val16, pcm_transition_silk); int pcm_transition_celt_size; VARDECL(opus_val16, pcm_transition_celt); opus_val16 *pcm_transition=NULL; int redundant_audio_size; VARDECL(opus_val16, redundant_audio); int audiosize; int mode; int transition=0; int start_band; int redundancy=0; int redundancy_bytes = 0; int celt_to_silk=0; int c; int F2_5, F5, F10, F20; const opus_val16 *window; opus_uint32 redundant_rng = 0; int celt_accum; ALLOC_STACK; silk_dec = (char*)st+st->silk_dec_offset; celt_dec = (CELTDecoder*)((char*)st+st->celt_dec_offset); F20 = st->Fs/50; F10 = F20>>1; F5 = F10>>1; F2_5 = F5>>1; if (frame_size < F2_5) { RESTORE_STACK; return OPUS_BUFFER_TOO_SMALL; } /* Limit frame_size to avoid excessive stack allocations. */ frame_size = IMIN(frame_size, st->Fs/25*3); /* Payloads of 1 (2 including ToC) or 0 trigger the PLC/DTX */ if (len<=1) { data = NULL; /* In that case, don't conceal more than what the ToC says */ frame_size = IMIN(frame_size, st->frame_size); } if (data != NULL) { audiosize = st->frame_size; mode = st->mode; ec_dec_init(&dec,(unsigned char*)data,len); } else { audiosize = frame_size; mode = st->prev_mode; if (mode == 0) { /* If we haven't got any packet yet, all we can do is return zeros */ for (i=0; i<audiosize*st->channels; i++) pcm[i] = 0; RESTORE_STACK; return audiosize; } /* Avoids trying to run the PLC on sizes other than 2.5 (CELT), 5 (CELT), 10, or 20 (e.g. 12.5 or 30 ms). */ if (audiosize > F20) { do { int ret = opus_decode_frame(st, NULL, 0, pcm, IMIN(audiosize, F20), 0); if (ret<0) { RESTORE_STACK; return ret; } pcm += ret*st->channels; audiosize -= ret; } while (audiosize > 0); RESTORE_STACK; return frame_size; } else if (audiosize < F20) { if (audiosize > F10) audiosize = F10; else if (mode != MODE_SILK_ONLY && audiosize > F5 && audiosize < F10) audiosize = F5; } } /* In fixed-point, we can tell CELT to do the accumulation on top of the SILK PCM buffer. This saves some stack space. */ #ifdef FIXED_POINT celt_accum = (mode != MODE_CELT_ONLY) && (frame_size >= F10); #else celt_accum = 0; #endif pcm_transition_silk_size = ALLOC_NONE; pcm_transition_celt_size = ALLOC_NONE; if (data!=NULL && st->prev_mode > 0 && ( (mode == MODE_CELT_ONLY && st->prev_mode != MODE_CELT_ONLY && !st->prev_redundancy) || (mode != MODE_CELT_ONLY && st->prev_mode == MODE_CELT_ONLY) ) ) { transition = 1; /* Decide where to allocate the stack memory for pcm_transition */ if (mode == MODE_CELT_ONLY) pcm_transition_celt_size = F5*st->channels; else pcm_transition_silk_size = F5*st->channels; } ALLOC(pcm_transition_celt, pcm_transition_celt_size, opus_val16); if (transition && mode == MODE_CELT_ONLY) { pcm_transition = pcm_transition_celt; opus_decode_frame(st, NULL, 0, pcm_transition, IMIN(F5, audiosize), 0); } if (audiosize > frame_size) { /*fprintf(stderr, "PCM buffer too small: %d vs %d (mode = %d)\n", audiosize, frame_size, mode);*/ RESTORE_STACK; return OPUS_BAD_ARG; } else { frame_size = audiosize; } /* Don't allocate any memory when in CELT-only mode */ pcm_silk_size = (mode != MODE_CELT_ONLY && !celt_accum) ? IMAX(F10, frame_size)*st->channels : ALLOC_NONE; ALLOC(pcm_silk, pcm_silk_size, opus_int16); /* SILK processing */ if (mode != MODE_CELT_ONLY) { int lost_flag, decoded_samples; opus_int16 *pcm_ptr; #ifdef FIXED_POINT if (celt_accum) pcm_ptr = pcm; else #endif pcm_ptr = pcm_silk; if (st->prev_mode==MODE_CELT_ONLY) silk_InitDecoder( silk_dec ); /* The SILK PLC cannot produce frames of less than 10 ms */ st->DecControl.payloadSize_ms = IMAX(10, 1000 * audiosize / st->Fs); if (data != NULL) { st->DecControl.nChannelsInternal = st->stream_channels; if( mode == MODE_SILK_ONLY ) { if( st->bandwidth == OPUS_BANDWIDTH_NARROWBAND ) { st->DecControl.internalSampleRate = 8000; } else if( st->bandwidth == OPUS_BANDWIDTH_MEDIUMBAND ) { st->DecControl.internalSampleRate = 12000; } else if( st->bandwidth == OPUS_BANDWIDTH_WIDEBAND ) { st->DecControl.internalSampleRate = 16000; } else { st->DecControl.internalSampleRate = 16000; silk_assert( 0 ); } } else { /* Hybrid mode */ st->DecControl.internalSampleRate = 16000; } } lost_flag = data == NULL ? 1 : 2 * decode_fec; decoded_samples = 0; do { /* Call SILK decoder */ int first_frame = decoded_samples == 0; silk_ret = silk_Decode( silk_dec, &st->DecControl, lost_flag, first_frame, &dec, pcm_ptr, &silk_frame_size ); if( silk_ret ) { if (lost_flag) { /* PLC failure should not be fatal */ silk_frame_size = frame_size; for (i=0; i<frame_size*st->channels; i++) pcm_ptr[i] = 0; } else { RESTORE_STACK; return OPUS_INTERNAL_ERROR; } } pcm_ptr += silk_frame_size * st->channels; decoded_samples += silk_frame_size; } while( decoded_samples < frame_size ); } start_band = 0; if (!decode_fec && mode != MODE_CELT_ONLY && data != NULL && ec_tell(&dec)+17+20*(st->mode == MODE_HYBRID) <= 8*len) { /* Check if we have a redundant 0-8 kHz band */ if (mode == MODE_HYBRID) redundancy = ec_dec_bit_logp(&dec, 12); else redundancy = 1; if (redundancy) { celt_to_silk = ec_dec_bit_logp(&dec, 1); /* redundancy_bytes will be at least two, in the non-hybrid case due to the ec_tell() check above */ redundancy_bytes = mode==MODE_HYBRID ? (opus_int32)ec_dec_uint(&dec, 256)+2 : len-((ec_tell(&dec)+7)>>3); len -= redundancy_bytes; /* This is a sanity check. It should never happen for a valid packet, so the exact behaviour is not normative. */ if (len*8 < ec_tell(&dec)) { len = 0; redundancy_bytes = 0; redundancy = 0; } /* Shrink decoder because of raw bits */ dec.storage -= redundancy_bytes; } } if (mode != MODE_CELT_ONLY) start_band = 17; { int endband=21; switch(st->bandwidth) { case OPUS_BANDWIDTH_NARROWBAND: endband = 13; break; case OPUS_BANDWIDTH_MEDIUMBAND: case OPUS_BANDWIDTH_WIDEBAND: endband = 17; break; case OPUS_BANDWIDTH_SUPERWIDEBAND: endband = 19; break; case OPUS_BANDWIDTH_FULLBAND: endband = 21; break; } celt_decoder_ctl(celt_dec, CELT_SET_END_BAND(endband)); celt_decoder_ctl(celt_dec, CELT_SET_CHANNELS(st->stream_channels)); } if (redundancy) { transition = 0; pcm_transition_silk_size=ALLOC_NONE; } ALLOC(pcm_transition_silk, pcm_transition_silk_size, opus_val16); if (transition && mode != MODE_CELT_ONLY) { pcm_transition = pcm_transition_silk; opus_decode_frame(st, NULL, 0, pcm_transition, IMIN(F5, audiosize), 0); } /* Only allocation memory for redundancy if/when needed */ redundant_audio_size = redundancy ? F5*st->channels : ALLOC_NONE; ALLOC(redundant_audio, redundant_audio_size, opus_val16); /* 5 ms redundant frame for CELT->SILK*/ if (redundancy && celt_to_silk) { celt_decoder_ctl(celt_dec, CELT_SET_START_BAND(0)); celt_decode_with_ec(celt_dec, data+len, redundancy_bytes, redundant_audio, F5, NULL, 0); celt_decoder_ctl(celt_dec, OPUS_GET_FINAL_RANGE(&redundant_rng)); } /* MUST be after PLC */ celt_decoder_ctl(celt_dec, CELT_SET_START_BAND(start_band)); if (mode != MODE_SILK_ONLY) { int celt_frame_size = IMIN(F20, frame_size); /* Make sure to discard any previous CELT state */ if (mode != st->prev_mode && st->prev_mode > 0 && !st->prev_redundancy) celt_decoder_ctl(celt_dec, OPUS_RESET_STATE); /* Decode CELT */ celt_ret = celt_decode_with_ec(celt_dec, decode_fec ? NULL : data, len, pcm, celt_frame_size, &dec, celt_accum); } else { unsigned char silence[2] = {0xFF, 0xFF}; if (!celt_accum) { for (i=0; i<frame_size*st->channels; i++) pcm[i] = 0; } /* For hybrid -> SILK transitions, we let the CELT MDCT do a fade-out by decoding a silence frame */ if (st->prev_mode == MODE_HYBRID && !(redundancy && celt_to_silk && st->prev_redundancy) ) { celt_decoder_ctl(celt_dec, CELT_SET_START_BAND(0)); celt_decode_with_ec(celt_dec, silence, 2, pcm, F2_5, NULL, celt_accum); } } if (mode != MODE_CELT_ONLY && !celt_accum) { #ifdef FIXED_POINT for (i=0; i<frame_size*st->channels; i++) pcm[i] = SAT16(ADD32(pcm[i], pcm_silk[i])); #else for (i=0; i<frame_size*st->channels; i++) pcm[i] = pcm[i] + (opus_val16)((1.f/32768.f)*pcm_silk[i]); #endif } { const CELTMode *celt_mode; celt_decoder_ctl(celt_dec, CELT_GET_MODE(&celt_mode)); window = celt_mode->window; } /* 5 ms redundant frame for SILK->CELT */ if (redundancy && !celt_to_silk) { celt_decoder_ctl(celt_dec, OPUS_RESET_STATE); celt_decoder_ctl(celt_dec, CELT_SET_START_BAND(0)); celt_decode_with_ec(celt_dec, data+len, redundancy_bytes, redundant_audio, F5, NULL, 0); celt_decoder_ctl(celt_dec, OPUS_GET_FINAL_RANGE(&redundant_rng)); smooth_fade(pcm+st->channels*(frame_size-F2_5), redundant_audio+st->channels*F2_5, pcm+st->channels*(frame_size-F2_5), F2_5, st->channels, window, st->Fs); } if (redundancy && celt_to_silk) { for (c=0; c<st->channels; c++) { for (i=0; i<F2_5; i++) pcm[st->channels*i+c] = redundant_audio[st->channels*i+c]; } smooth_fade(redundant_audio+st->channels*F2_5, pcm+st->channels*F2_5, pcm+st->channels*F2_5, F2_5, st->channels, window, st->Fs); } if (transition) { if (audiosize >= F5) { for (i=0; i<st->channels*F2_5; i++) pcm[i] = pcm_transition[i]; smooth_fade(pcm_transition+st->channels*F2_5, pcm+st->channels*F2_5, pcm+st->channels*F2_5, F2_5, st->channels, window, st->Fs); } else { /* Not enough time to do a clean transition, but we do it anyway This will not preserve amplitude perfectly and may introduce a bit of temporal aliasing, but it shouldn't be too bad and that's pretty much the best we can do. In any case, generating this transition it pretty silly in the first place */ smooth_fade(pcm_transition, pcm, pcm, F2_5, st->channels, window, st->Fs); } } if(st->decode_gain) { opus_val32 gain; gain = celt_exp2(MULT16_16_P15(QCONST16(6.48814081e-4f, 25), st->decode_gain)); for (i=0; i<frame_size*st->channels; i++) { opus_val32 x; x = MULT16_32_P16(pcm[i],gain); pcm[i] = SATURATE(x, 32767); } } if (len <= 1) st->rangeFinal = 0; else st->rangeFinal = dec.rng ^ redundant_rng; st->prev_mode = mode; st->prev_redundancy = redundancy && !celt_to_silk; if (celt_ret>=0) { if (OPUS_CHECK_ARRAY(pcm, audiosize*st->channels)) OPUS_PRINT_INT(audiosize); } RESTORE_STACK; return celt_ret < 0 ? celt_ret : audiosize; }
static int opus_decode_frame(OpusStreamContext *s, const uint8_t *data, int size) { int samples = s->packet.frame_duration; int redundancy = 0; int redundancy_size, redundancy_pos; int ret, i, consumed; int delayed_samples = s->delayed_samples; ret = ff_opus_rc_dec_init(&s->rc, data, size); if (ret < 0) return ret; /* decode the silk frame */ if (s->packet.mode == OPUS_MODE_SILK || s->packet.mode == OPUS_MODE_HYBRID) { #if CONFIG_SWRESAMPLE if (!swr_is_initialized(s->swr)) { #elif CONFIG_AVRESAMPLE if (!avresample_is_open(s->avr)) { #endif ret = opus_init_resample(s); if (ret < 0) return ret; } samples = ff_silk_decode_superframe(s->silk, &s->rc, s->silk_output, FFMIN(s->packet.bandwidth, OPUS_BANDWIDTH_WIDEBAND), s->packet.stereo + 1, silk_frame_duration_ms[s->packet.config]); if (samples < 0) { av_log(s->avctx, AV_LOG_ERROR, "Error decoding a SILK frame.\n"); return samples; } #if CONFIG_SWRESAMPLE samples = swr_convert(s->swr, (uint8_t**)s->out, s->packet.frame_duration, (const uint8_t**)s->silk_output, samples); #elif CONFIG_AVRESAMPLE samples = avresample_convert(s->avr, (uint8_t**)s->out, s->out_size, s->packet.frame_duration, (uint8_t**)s->silk_output, sizeof(s->silk_buf[0]), samples); #endif if (samples < 0) { av_log(s->avctx, AV_LOG_ERROR, "Error resampling SILK data.\n"); return samples; } av_assert2((samples & 7) == 0); s->delayed_samples += s->packet.frame_duration - samples; } else ff_silk_flush(s->silk); // decode redundancy information consumed = opus_rc_tell(&s->rc); if (s->packet.mode == OPUS_MODE_HYBRID && consumed + 37 <= size * 8) redundancy = ff_opus_rc_dec_log(&s->rc, 12); else if (s->packet.mode == OPUS_MODE_SILK && consumed + 17 <= size * 8) redundancy = 1; if (redundancy) { redundancy_pos = ff_opus_rc_dec_log(&s->rc, 1); if (s->packet.mode == OPUS_MODE_HYBRID) redundancy_size = ff_opus_rc_dec_uint(&s->rc, 256) + 2; else redundancy_size = size - (consumed + 7) / 8; size -= redundancy_size; if (size < 0) { av_log(s->avctx, AV_LOG_ERROR, "Invalid redundancy frame size.\n"); return AVERROR_INVALIDDATA; } if (redundancy_pos) { ret = opus_decode_redundancy(s, data + size, redundancy_size); if (ret < 0) return ret; ff_celt_flush(s->celt); } } /* decode the CELT frame */ if (s->packet.mode == OPUS_MODE_CELT || s->packet.mode == OPUS_MODE_HYBRID) { float *out_tmp[2] = { s->out[0], s->out[1] }; float **dst = (s->packet.mode == OPUS_MODE_CELT) ? out_tmp : s->celt_output; int celt_output_samples = samples; int delay_samples = av_audio_fifo_size(s->celt_delay); if (delay_samples) { if (s->packet.mode == OPUS_MODE_HYBRID) { av_audio_fifo_read(s->celt_delay, (void**)s->celt_output, delay_samples); for (i = 0; i < s->output_channels; i++) { s->fdsp->vector_fmac_scalar(out_tmp[i], s->celt_output[i], 1.0, delay_samples); out_tmp[i] += delay_samples; } celt_output_samples -= delay_samples; } else { av_log(s->avctx, AV_LOG_WARNING, "Spurious CELT delay samples present.\n"); av_audio_fifo_drain(s->celt_delay, delay_samples); if (s->avctx->err_recognition & AV_EF_EXPLODE) return AVERROR_BUG; } } ff_opus_rc_dec_raw_init(&s->rc, data + size, size); ret = ff_celt_decode_frame(s->celt, &s->rc, dst, s->packet.stereo + 1, s->packet.frame_duration, (s->packet.mode == OPUS_MODE_HYBRID) ? 17 : 0, ff_celt_band_end[s->packet.bandwidth]); if (ret < 0) return ret; if (s->packet.mode == OPUS_MODE_HYBRID) { int celt_delay = s->packet.frame_duration - celt_output_samples; void *delaybuf[2] = { s->celt_output[0] + celt_output_samples, s->celt_output[1] + celt_output_samples }; for (i = 0; i < s->output_channels; i++) { s->fdsp->vector_fmac_scalar(out_tmp[i], s->celt_output[i], 1.0, celt_output_samples); } ret = av_audio_fifo_write(s->celt_delay, delaybuf, celt_delay); if (ret < 0) return ret; } } else ff_celt_flush(s->celt); if (s->redundancy_idx) { for (i = 0; i < s->output_channels; i++) opus_fade(s->out[i], s->out[i], s->redundancy_output[i] + 120 + s->redundancy_idx, ff_celt_window2 + s->redundancy_idx, 120 - s->redundancy_idx); s->redundancy_idx = 0; } if (redundancy) { if (!redundancy_pos) { ff_celt_flush(s->celt); ret = opus_decode_redundancy(s, data + size, redundancy_size); if (ret < 0) return ret; for (i = 0; i < s->output_channels; i++) { opus_fade(s->out[i] + samples - 120 + delayed_samples, s->out[i] + samples - 120 + delayed_samples, s->redundancy_output[i] + 120, ff_celt_window2, 120 - delayed_samples); if (delayed_samples) s->redundancy_idx = 120 - delayed_samples; } } else { for (i = 0; i < s->output_channels; i++) { memcpy(s->out[i] + delayed_samples, s->redundancy_output[i], 120 * sizeof(float)); opus_fade(s->out[i] + 120 + delayed_samples, s->redundancy_output[i] + 120, s->out[i] + 120 + delayed_samples, ff_celt_window2, 120); } } } return samples; } static int opus_decode_subpacket(OpusStreamContext *s, const uint8_t *buf, int buf_size, float **out, int out_size, int nb_samples) { int output_samples = 0; int flush_needed = 0; int i, j, ret; s->out[0] = out[0]; s->out[1] = out[1]; s->out_size = out_size; /* check if we need to flush the resampler */ #if CONFIG_SWRESAMPLE if (swr_is_initialized(s->swr)) { if (buf) { int64_t cur_samplerate; av_opt_get_int(s->swr, "in_sample_rate", 0, &cur_samplerate); flush_needed = (s->packet.mode == OPUS_MODE_CELT) || (cur_samplerate != s->silk_samplerate); } else { flush_needed = !!s->delayed_samples; } } #elif CONFIG_AVRESAMPLE if (avresample_is_open(s->avr)) { if (buf) { int64_t cur_samplerate; av_opt_get_int(s->avr, "in_sample_rate", 0, &cur_samplerate); flush_needed = (s->packet.mode == OPUS_MODE_CELT) || (cur_samplerate != s->silk_samplerate); } else { flush_needed = !!s->delayed_samples; } } #endif if (!buf && !flush_needed) return 0; /* use dummy output buffers if the channel is not mapped to anything */ if (!s->out[0] || (s->output_channels == 2 && !s->out[1])) { av_fast_malloc(&s->out_dummy, &s->out_dummy_allocated_size, s->out_size); if (!s->out_dummy) return AVERROR(ENOMEM); if (!s->out[0]) s->out[0] = s->out_dummy; if (!s->out[1]) s->out[1] = s->out_dummy; } /* flush the resampler if necessary */ if (flush_needed) { ret = opus_flush_resample(s, s->delayed_samples); if (ret < 0) { av_log(s->avctx, AV_LOG_ERROR, "Error flushing the resampler.\n"); return ret; } #if CONFIG_SWRESAMPLE swr_close(s->swr); #elif CONFIG_AVRESAMPLE avresample_close(s->avr); #endif output_samples += s->delayed_samples; s->delayed_samples = 0; if (!buf) goto finish; } /* decode all the frames in the packet */ for (i = 0; i < s->packet.frame_count; i++) { int size = s->packet.frame_size[i]; int samples = opus_decode_frame(s, buf + s->packet.frame_offset[i], size); if (samples < 0) { av_log(s->avctx, AV_LOG_ERROR, "Error decoding an Opus frame.\n"); if (s->avctx->err_recognition & AV_EF_EXPLODE) return samples; for (j = 0; j < s->output_channels; j++) memset(s->out[j], 0, s->packet.frame_duration * sizeof(float)); samples = s->packet.frame_duration; } output_samples += samples; for (j = 0; j < s->output_channels; j++) s->out[j] += samples; s->out_size -= samples * sizeof(float); } finish: s->out[0] = s->out[1] = NULL; s->out_size = 0; return output_samples; }
static int opus_decode_subpacket(OpusStreamContext *s, const uint8_t *buf, int buf_size, int nb_samples) { int output_samples = 0; int flush_needed = 0; int i, j, ret; /* check if we need to flush the resampler */ if (swr_is_initialized(s->swr)) { if (buf) { int64_t cur_samplerate; av_opt_get_int(s->swr, "in_sample_rate", 0, &cur_samplerate); flush_needed = (s->packet.mode == OPUS_MODE_CELT) || (cur_samplerate != s->silk_samplerate); } else { flush_needed = !!s->delayed_samples; } } if (!buf && !flush_needed) return 0; /* use dummy output buffers if the channel is not mapped to anything */ if (!s->out[0] || (s->output_channels == 2 && !s->out[1])) { av_fast_malloc(&s->out_dummy, &s->out_dummy_allocated_size, s->out_size); if (!s->out_dummy) return AVERROR(ENOMEM); if (!s->out[0]) s->out[0] = s->out_dummy; if (!s->out[1]) s->out[1] = s->out_dummy; } /* flush the resampler if necessary */ if (flush_needed) { ret = opus_flush_resample(s, s->delayed_samples); if (ret < 0) { av_log(s->avctx, AV_LOG_ERROR, "Error flushing the resampler.\n"); return ret; } swr_close(s->swr); output_samples += s->delayed_samples; s->delayed_samples = 0; if (!buf) goto finish; } /* decode all the frames in the packet */ for (i = 0; i < s->packet.frame_count; i++) { int size = s->packet.frame_size[i]; int samples = opus_decode_frame(s, buf + s->packet.frame_offset[i], size); if (samples < 0) { av_log(s->avctx, AV_LOG_ERROR, "Error decoding an Opus frame.\n"); if (s->avctx->err_recognition & AV_EF_EXPLODE) return samples; for (j = 0; j < s->output_channels; j++) memset(s->out[j], 0, s->packet.frame_duration * sizeof(float)); samples = s->packet.frame_duration; } output_samples += samples; for (j = 0; j < s->output_channels; j++) s->out[j] += samples; s->out_size -= samples * sizeof(float); } finish: s->out[0] = s->out[1] = NULL; s->out_size = 0; return output_samples; }
int opus_decode_native(OpusDecoder *st, const unsigned char *data, opus_int32 len, opus_val16 *pcm, int frame_size, int decode_fec, int self_delimited, int *packet_offset) { int i, nb_samples; int count, offset; unsigned char toc; int tot_offset; int packet_frame_size, packet_bandwidth, packet_mode, packet_stream_channels; /* 48 x 2.5 ms = 120 ms */ short size[48]; if (decode_fec<0 || decode_fec>1) return OPUS_BAD_ARG; /* For FEC/PLC, frame_size has to be to have a multiple of 2.5 ms */ if ((decode_fec || len==0 || data==NULL) && frame_size%(st->Fs/400)!=0) return OPUS_BAD_ARG; if (len==0 || data==NULL) { int pcm_count=0; do { int ret; ret = opus_decode_frame(st, NULL, 0, pcm, frame_size-pcm_count, 0); if (ret<0) return ret; pcm += st->channels*ret; pcm_count += ret; } while (pcm_count < frame_size); st->last_packet_duration = pcm_count; return pcm_count; } else if (len<0) return OPUS_BAD_ARG; packet_mode = opus_packet_get_mode(data); packet_bandwidth = opus_packet_get_bandwidth(data); packet_frame_size = opus_packet_get_samples_per_frame(data, st->Fs); packet_stream_channels = opus_packet_get_nb_channels(data); count = opus_packet_parse_impl(data, len, self_delimited, &toc, NULL, size, &offset); data += offset; if (decode_fec) { int duration_copy; int ret; /* If no FEC can be present, run the PLC (recursive call) */ if (frame_size <= packet_frame_size || packet_mode == MODE_CELT_ONLY || st->mode == MODE_CELT_ONLY) return opus_decode_native(st, NULL, 0, pcm, frame_size, 0, 0, NULL); /* Otherwise, run the PLC on everything except the size for which we might have FEC */ duration_copy = st->last_packet_duration; ret = opus_decode_native(st, NULL, 0, pcm, frame_size-packet_frame_size, 0, 0, NULL); if (ret<0) { st->last_packet_duration = duration_copy; return ret; } /* Complete with FEC */ st->mode = packet_mode; st->bandwidth = packet_bandwidth; st->frame_size = packet_frame_size; st->stream_channels = packet_stream_channels; ret = opus_decode_frame(st, data, size[0], pcm+st->channels*(frame_size-packet_frame_size), packet_frame_size, 1); if (ret<0) return ret; st->last_packet_duration = frame_size; return frame_size; } tot_offset = 0; if (count < 0) return count; tot_offset += offset; if (count*packet_frame_size > frame_size) return OPUS_BUFFER_TOO_SMALL; /* Update the state as the last step to avoid updating it on an invalid packet */ st->mode = packet_mode; st->bandwidth = packet_bandwidth; st->frame_size = packet_frame_size; st->stream_channels = packet_stream_channels; nb_samples=0; for (i=0;i<count;i++) { int ret; ret = opus_decode_frame(st, data, size[i], pcm, frame_size-nb_samples, decode_fec); if (ret<0) return ret; data += size[i]; tot_offset += size[i]; pcm += ret*st->channels; nb_samples += ret; } if (packet_offset != NULL) *packet_offset = tot_offset; st->last_packet_duration = nb_samples; return nb_samples; }