static int cng_decode_frame(AVCodecContext *avctx, void *data, int *got_frame_ptr, AVPacket *avpkt) { AVFrame *frame = data; CNGContext *p = avctx->priv_data; int buf_size = avpkt->size; int ret, i; int16_t *buf_out; float e = 1.0; float scaling; if (avpkt->size) { int dbov = -avpkt->data[0]; p->target_energy = 1081109975 * ff_exp10(dbov / 10.0) * 0.75; memset(p->target_refl_coef, 0, p->order * sizeof(*p->target_refl_coef)); for (i = 0; i < FFMIN(avpkt->size - 1, p->order); i++) { p->target_refl_coef[i] = (avpkt->data[1 + i] - 127) / 128.0; } } if (p->inited) { p->energy = p->energy / 2 + p->target_energy / 2; for (i = 0; i < p->order; i++) p->refl_coef[i] = 0.6 *p->refl_coef[i] + 0.4 * p->target_refl_coef[i]; } else { p->energy = p->target_energy; memcpy(p->refl_coef, p->target_refl_coef, p->order * sizeof(*p->refl_coef)); p->inited = 1; } make_lpc_coefs(p->lpc_coef, p->refl_coef, p->order); for (i = 0; i < p->order; i++) e *= 1.0 - p->refl_coef[i]*p->refl_coef[i]; scaling = sqrt(e * p->energy / 1081109975); for (i = 0; i < avctx->frame_size; i++) { int r = (av_lfg_get(&p->lfg) & 0xffff) - 0x8000; p->excitation[i] = scaling * r; } ff_celp_lp_synthesis_filterf(p->filter_out + p->order, p->lpc_coef, p->excitation, avctx->frame_size, p->order); frame->nb_samples = avctx->frame_size; if ((ret = ff_get_buffer(avctx, frame, 0)) < 0) return ret; buf_out = (int16_t *)frame->data[0]; for (i = 0; i < avctx->frame_size; i++) buf_out[i] = av_clip_int16(p->filter_out[i + p->order]); memcpy(p->filter_out, p->filter_out + avctx->frame_size, p->order * sizeof(*p->filter_out)); *got_frame_ptr = 1; return buf_size; }
// FIXME use for decoding too static void init_exp(WMACodecContext *s, int ch, const int *exp_param) { int n; const uint16_t *ptr; float v, *q, max_scale, *q_end; ptr = s->exponent_bands[s->frame_len_bits - s->block_len_bits]; q = s->exponents[ch]; q_end = q + s->block_len; max_scale = 0; while (q < q_end) { /* XXX: use a table */ v = ff_exp10(*exp_param++ *(1.0 / 16.0)); max_scale = FFMAX(max_scale, v); n = *ptr++; do { *q++ = v; } while (--n); } s->max_exponent[ch] = max_scale; }
static void read8_fft_bin(SpectrumSynthContext *s, int x, int y, int f, int ch) { const int m_linesize = s->magnitude->linesize[0]; const int p_linesize = s->phase->linesize[0]; const uint8_t *m = (uint8_t *)(s->magnitude->data[0] + y * m_linesize); const uint8_t *p = (uint8_t *)(s->phase->data[0] + y * p_linesize); float magnitude, phase; switch (s->scale) { case LINEAR: magnitude = m[x] / (double)UINT8_MAX; break; case LOG: magnitude = ff_exp10(((m[x] / (double)UINT8_MAX) - 1.) * 6.); break; default: av_assert0(0); } phase = ((p[x] / (double)UINT8_MAX) * 2. - 1.) * M_PI; s->fft_data[ch][f].re = magnitude * cos(phase); s->fft_data[ch][f].im = magnitude * sin(phase); }
static int encode_block(WMACodecContext *s, float (*src_coefs)[BLOCK_MAX_SIZE], int total_gain) { int v, bsize, ch, coef_nb_bits, parse_exponents; float mdct_norm; int nb_coefs[MAX_CHANNELS]; static const int fixed_exp[25] = { 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20 }; // FIXME remove duplication relative to decoder if (s->use_variable_block_len) { av_assert0(0); // FIXME not implemented } else { /* fixed block len */ s->next_block_len_bits = s->frame_len_bits; s->prev_block_len_bits = s->frame_len_bits; s->block_len_bits = s->frame_len_bits; } s->block_len = 1 << s->block_len_bits; // av_assert0((s->block_pos + s->block_len) <= s->frame_len); bsize = s->frame_len_bits - s->block_len_bits; // FIXME factor v = s->coefs_end[bsize] - s->coefs_start; for (ch = 0; ch < s->avctx->channels; ch++) nb_coefs[ch] = v; { int n4 = s->block_len / 2; mdct_norm = 1.0 / (float) n4; if (s->version == 1) mdct_norm *= sqrt(n4); } if (s->avctx->channels == 2) put_bits(&s->pb, 1, !!s->ms_stereo); for (ch = 0; ch < s->avctx->channels; ch++) { // FIXME only set channel_coded when needed, instead of always s->channel_coded[ch] = 1; if (s->channel_coded[ch]) init_exp(s, ch, fixed_exp); } for (ch = 0; ch < s->avctx->channels; ch++) { if (s->channel_coded[ch]) { WMACoef *coefs1; float *coefs, *exponents, mult; int i, n; coefs1 = s->coefs1[ch]; exponents = s->exponents[ch]; mult = ff_exp10(total_gain * 0.05) / s->max_exponent[ch]; mult *= mdct_norm; coefs = src_coefs[ch]; if (s->use_noise_coding && 0) { av_assert0(0); // FIXME not implemented } else { coefs += s->coefs_start; n = nb_coefs[ch]; for (i = 0; i < n; i++) { double t = *coefs++ / (exponents[i] * mult); if (t < -32768 || t > 32767) return -1; coefs1[i] = lrint(t); } } } } v = 0; for (ch = 0; ch < s->avctx->channels; ch++) { int a = s->channel_coded[ch]; put_bits(&s->pb, 1, a); v |= a; } if (!v) return 1; for (v = total_gain - 1; v >= 127; v -= 127) put_bits(&s->pb, 7, 127); put_bits(&s->pb, 7, v); coef_nb_bits = ff_wma_total_gain_to_bits(total_gain); if (s->use_noise_coding) { for (ch = 0; ch < s->avctx->channels; ch++) { if (s->channel_coded[ch]) { int i, n; n = s->exponent_high_sizes[bsize]; for (i = 0; i < n; i++) { put_bits(&s->pb, 1, s->high_band_coded[ch][i] = 0); if (0) nb_coefs[ch] -= s->exponent_high_bands[bsize][i]; } } } } parse_exponents = 1; if (s->block_len_bits != s->frame_len_bits) put_bits(&s->pb, 1, parse_exponents); if (parse_exponents) { for (ch = 0; ch < s->avctx->channels; ch++) { if (s->channel_coded[ch]) { if (s->use_exp_vlc) { encode_exp_vlc(s, ch, fixed_exp); } else { av_assert0(0); // FIXME not implemented // encode_exp_lsp(s, ch); } } } } else av_assert0(0); // FIXME not implemented for (ch = 0; ch < s->avctx->channels; ch++) { if (s->channel_coded[ch]) { int run, tindex; WMACoef *ptr, *eptr; tindex = (ch == 1 && s->ms_stereo); ptr = &s->coefs1[ch][0]; eptr = ptr + nb_coefs[ch]; run = 0; for (; ptr < eptr; ptr++) { if (*ptr) { int level = *ptr; int abs_level = FFABS(level); int code = 0; if (abs_level <= s->coef_vlcs[tindex]->max_level) if (run < s->coef_vlcs[tindex]->levels[abs_level - 1]) code = run + s->int_table[tindex][abs_level - 1]; av_assert2(code < s->coef_vlcs[tindex]->n); put_bits(&s->pb, s->coef_vlcs[tindex]->huffbits[code], s->coef_vlcs[tindex]->huffcodes[code]); if (code == 0) { if (1 << coef_nb_bits <= abs_level) return -1; put_bits(&s->pb, coef_nb_bits, abs_level); put_bits(&s->pb, s->frame_len_bits, run); } // FIXME the sign is flipped somewhere put_bits(&s->pb, 1, level < 0); run = 0; } else run++; } if (run) put_bits(&s->pb, s->coef_vlcs[tindex]->huffbits[1], s->coef_vlcs[tindex]->huffcodes[1]); } if (s->version == 1 && s->avctx->channels >= 2) avpriv_align_put_bits(&s->pb); } return 0; }
static int encode_init(AVCodecContext *avctx) { DCAEncContext *c = avctx->priv_data; uint64_t layout = avctx->channel_layout; int i, min_frame_bits; c->fullband_channels = c->channels = avctx->channels; c->lfe_channel = (avctx->channels == 3 || avctx->channels == 6); c->band_interpolation = band_interpolation[1]; c->band_spectrum = band_spectrum[1]; c->worst_quantization_noise = -2047; c->worst_noise_ever = -2047; if (!layout) { av_log(avctx, AV_LOG_WARNING, "No channel layout specified. The " "encoder will guess the layout, but it " "might be incorrect.\n"); layout = av_get_default_channel_layout(avctx->channels); } switch (layout) { case AV_CH_LAYOUT_MONO: c->channel_config = 0; break; case AV_CH_LAYOUT_STEREO: c->channel_config = 2; break; case AV_CH_LAYOUT_2_2: c->channel_config = 8; break; case AV_CH_LAYOUT_5POINT0: c->channel_config = 9; break; case AV_CH_LAYOUT_5POINT1: c->channel_config = 9; break; default: av_log(avctx, AV_LOG_ERROR, "Unsupported channel layout!\n"); return AVERROR_PATCHWELCOME; } if (c->lfe_channel) { c->fullband_channels--; c->channel_order_tab = ff_dca_channel_reorder_lfe[c->channel_config]; } else { c->channel_order_tab = ff_dca_channel_reorder_nolfe[c->channel_config]; } for (i = 0; i < 9; i++) { if (sample_rates[i] == avctx->sample_rate) break; } if (i == 9) return AVERROR(EINVAL); c->samplerate_index = i; if (avctx->bit_rate < 32000 || avctx->bit_rate > 3840000) { av_log(avctx, AV_LOG_ERROR, "Bit rate %"PRId64" not supported.", (int64_t)avctx->bit_rate); return AVERROR(EINVAL); } for (i = 0; ff_dca_bit_rates[i] < avctx->bit_rate; i++) ; c->bitrate_index = i; c->frame_bits = FFALIGN((avctx->bit_rate * 512 + avctx->sample_rate - 1) / avctx->sample_rate, 32); min_frame_bits = 132 + (493 + 28 * 32) * c->fullband_channels + c->lfe_channel * 72; if (c->frame_bits < min_frame_bits || c->frame_bits > (DCA_MAX_FRAME_SIZE << 3)) return AVERROR(EINVAL); c->frame_size = (c->frame_bits + 7) / 8; avctx->frame_size = 32 * SUBBAND_SAMPLES; if (!cos_table[0]) { int j, k; cos_table[0] = 0x7fffffff; cos_table[512] = 0; cos_table[1024] = -cos_table[0]; for (i = 1; i < 512; i++) { cos_table[i] = (int32_t)(0x7fffffff * cos(M_PI * i / 1024)); cos_table[1024-i] = -cos_table[i]; cos_table[1024+i] = -cos_table[i]; cos_table[2048-i] = cos_table[i]; } for (i = 0; i < 2048; i++) { cb_to_level[i] = (int32_t)(0x7fffffff * ff_exp10(-0.005 * i)); } for (k = 0; k < 32; k++) { for (j = 0; j < 8; j++) { lfe_fir_64i[64 * j + k] = (int32_t)(0xffffff800000ULL * ff_dca_lfe_fir_64[8 * k + j]); lfe_fir_64i[64 * (7-j) + (63 - k)] = (int32_t)(0xffffff800000ULL * ff_dca_lfe_fir_64[8 * k + j]); } } for (i = 0; i < 512; i++) { band_interpolation[0][i] = (int32_t)(0x1000000000ULL * ff_dca_fir_32bands_perfect[i]); band_interpolation[1][i] = (int32_t)(0x1000000000ULL * ff_dca_fir_32bands_nonperfect[i]); } for (i = 0; i < 9; i++) { for (j = 0; j < AUBANDS; j++) { for (k = 0; k < 256; k++) { double freq = sample_rates[i] * (k + 0.5) / 512; auf[i][j][k] = (int32_t)(10 * (hom(freq) + gammafilter(j, freq))); } } } for (i = 0; i < 256; i++) { double add = 1 + ff_exp10(-0.01 * i); cb_to_add[i] = (int32_t)(100 * log10(add)); } for (j = 0; j < 8; j++) { double accum = 0; for (i = 0; i < 512; i++) { double reconst = ff_dca_fir_32bands_perfect[i] * ((i & 64) ? (-1) : 1); accum += reconst * cos(2 * M_PI * (i + 0.5 - 256) * (j + 0.5) / 512); } band_spectrum[0][j] = (int32_t)(200 * log10(accum)); } for (j = 0; j < 8; j++) { double accum = 0; for (i = 0; i < 512; i++) { double reconst = ff_dca_fir_32bands_nonperfect[i] * ((i & 64) ? (-1) : 1); accum += reconst * cos(2 * M_PI * (i + 0.5 - 256) * (j + 0.5) / 512); } band_spectrum[1][j] = (int32_t)(200 * log10(accum)); } } return 0; }