FFmpeg coverage

Directory: ../../../ffmpeg/
File: src/libavcodec/aacpsy.c
Date: 2025-11-10 16:28:23
Exec Total Coverage
Lines: 372 455 81.8%
Functions: 16 18 88.9%
Branches: 266 349 76.2%
Line Branch Exec Source
1 /*
2 * AAC encoder psychoacoustic model
3 * Copyright (C) 2008 Konstantin Shishkov
4 *
5 * This file is part of FFmpeg.
6 *
7 * FFmpeg is free software; you can redistribute it and/or
8 * modify it under the terms of the GNU Lesser General Public
9 * License as published by the Free Software Foundation; either
10 * version 2.1 of the License, or (at your option) any later version.
11 *
12 * FFmpeg is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 * Lesser General Public License for more details.
16 *
17 * You should have received a copy of the GNU Lesser General Public
18 * License along with FFmpeg; if not, write to the Free Software
19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
20 */
21
22 /**
23 * @file
24 * AAC encoder psychoacoustic model
25 */
26
27 #include "libavutil/attributes.h"
28 #include "libavutil/ffmath.h"
29 #include "libavutil/mem.h"
30
31 #include "avcodec.h"
32 #include "aac.h"
33 #include "psymodel.h"
34
35 /***********************************
36 * TODOs:
37 * try other bitrate controlling mechanism (maybe use ratecontrol.c?)
38 * control quality for quality-based output
39 **********************************/
40
41 /**
42 * constants for 3GPP AAC psychoacoustic model
43 * @{
44 */
45 #define PSY_3GPP_THR_SPREAD_HI 1.5f // spreading factor for low-to-hi threshold spreading (15 dB/Bark)
46 #define PSY_3GPP_THR_SPREAD_LOW 3.0f // spreading factor for hi-to-low threshold spreading (30 dB/Bark)
47 /* spreading factor for low-to-hi energy spreading, long block, > 22kbps/channel (20dB/Bark) */
48 #define PSY_3GPP_EN_SPREAD_HI_L1 2.0f
49 /* spreading factor for low-to-hi energy spreading, long block, <= 22kbps/channel (15dB/Bark) */
50 #define PSY_3GPP_EN_SPREAD_HI_L2 1.5f
51 /* spreading factor for low-to-hi energy spreading, short block (15 dB/Bark) */
52 #define PSY_3GPP_EN_SPREAD_HI_S 1.5f
53 /* spreading factor for hi-to-low energy spreading, long block (30dB/Bark) */
54 #define PSY_3GPP_EN_SPREAD_LOW_L 3.0f
55 /* spreading factor for hi-to-low energy spreading, short block (20dB/Bark) */
56 #define PSY_3GPP_EN_SPREAD_LOW_S 2.0f
57
58 #define PSY_3GPP_RPEMIN 0.01f
59 #define PSY_3GPP_RPELEV 2.0f
60
61 #define PSY_3GPP_C1 3.0f /* log2(8) */
62 #define PSY_3GPP_C2 1.3219281f /* log2(2.5) */
63 #define PSY_3GPP_C3 0.55935729f /* 1 - C2 / C1 */
64
65 #define PSY_SNR_1DB 7.9432821e-1f /* -1dB */
66 #define PSY_SNR_25DB 3.1622776e-3f /* -25dB */
67
68 #define PSY_3GPP_SAVE_SLOPE_L -0.46666667f
69 #define PSY_3GPP_SAVE_SLOPE_S -0.36363637f
70 #define PSY_3GPP_SAVE_ADD_L -0.84285712f
71 #define PSY_3GPP_SAVE_ADD_S -0.75f
72 #define PSY_3GPP_SPEND_SLOPE_L 0.66666669f
73 #define PSY_3GPP_SPEND_SLOPE_S 0.81818181f
74 #define PSY_3GPP_SPEND_ADD_L -0.35f
75 #define PSY_3GPP_SPEND_ADD_S -0.26111111f
76 #define PSY_3GPP_CLIP_LO_L 0.2f
77 #define PSY_3GPP_CLIP_LO_S 0.2f
78 #define PSY_3GPP_CLIP_HI_L 0.95f
79 #define PSY_3GPP_CLIP_HI_S 0.75f
80
81 #define PSY_3GPP_AH_THR_LONG 0.5f
82 #define PSY_3GPP_AH_THR_SHORT 0.63f
83
84 #define PSY_PE_FORGET_SLOPE 511
85
86 enum {
87 PSY_3GPP_AH_NONE,
88 PSY_3GPP_AH_INACTIVE,
89 PSY_3GPP_AH_ACTIVE
90 };
91
92 #define PSY_3GPP_BITS_TO_PE(bits) ((bits) * 1.18f)
93 #define PSY_3GPP_PE_TO_BITS(bits) ((bits) / 1.18f)
94
95 /* LAME psy model constants */
96 #define PSY_LAME_FIR_LEN 21 ///< LAME psy model FIR order
97 #define AAC_BLOCK_SIZE_LONG 1024 ///< long block size
98 #define AAC_BLOCK_SIZE_SHORT 128 ///< short block size
99 #define AAC_NUM_BLOCKS_SHORT 8 ///< number of blocks in a short sequence
100 #define PSY_LAME_NUM_SUBBLOCKS 2 ///< Number of sub-blocks in each short block
101
102 /**
103 * @}
104 */
105
106 /**
107 * information for single band used by 3GPP TS26.403-inspired psychoacoustic model
108 */
109 typedef struct AacPsyBand{
110 float energy; ///< band energy
111 float thr; ///< energy threshold
112 float thr_quiet; ///< threshold in quiet
113 float nz_lines; ///< number of non-zero spectral lines
114 float active_lines; ///< number of active spectral lines
115 float pe; ///< perceptual entropy
116 float pe_const; ///< constant part of the PE calculation
117 float norm_fac; ///< normalization factor for linearization
118 int avoid_holes; ///< hole avoidance flag
119 }AacPsyBand;
120
121 /**
122 * single/pair channel context for psychoacoustic model
123 */
124 typedef struct AacPsyChannel{
125 AacPsyBand band[128]; ///< bands information
126 AacPsyBand prev_band[128]; ///< bands information from the previous frame
127
128 float win_energy; ///< sliding average of channel energy
129 float iir_state[2]; ///< hi-pass IIR filter state
130 uint8_t next_grouping; ///< stored grouping scheme for the next frame (in case of 8 short window sequence)
131 enum WindowSequence next_window_seq; ///< window sequence to be used in the next frame
132 /* LAME psy model specific members */
133 float attack_threshold; ///< attack threshold for this channel
134 float prev_energy_subshort[AAC_NUM_BLOCKS_SHORT * PSY_LAME_NUM_SUBBLOCKS];
135 int prev_attack; ///< attack value for the last short block in the previous sequence
136 int next_attack0_zero; ///< whether attack[0] of the next frame is zero
137 }AacPsyChannel;
138
139 /**
140 * psychoacoustic model frame type-dependent coefficients
141 */
142 typedef struct AacPsyCoeffs{
143 float ath; ///< absolute threshold of hearing per bands
144 float barks; ///< Bark value for each spectral band in long frame
145 float spread_low[2]; ///< spreading factor for low-to-high threshold spreading in long frame
146 float spread_hi [2]; ///< spreading factor for high-to-low threshold spreading in long frame
147 float min_snr; ///< minimal SNR
148 }AacPsyCoeffs;
149
150 /**
151 * 3GPP TS26.403-inspired psychoacoustic model specific data
152 */
153 typedef struct AacPsyContext{
154 int chan_bitrate; ///< bitrate per channel
155 int frame_bits; ///< average bits per frame
156 int fill_level; ///< bit reservoir fill level
157 struct {
158 float min; ///< minimum allowed PE for bit factor calculation
159 float max; ///< maximum allowed PE for bit factor calculation
160 float previous; ///< allowed PE of the previous frame
161 float correction; ///< PE correction factor
162 } pe;
163 AacPsyCoeffs psy_coef[2][64];
164 AacPsyChannel *ch;
165 float global_quality; ///< normalized global quality taken from avctx
166 }AacPsyContext;
167
168 /**
169 * LAME psy model preset struct
170 */
171 typedef struct PsyLamePreset {
172 int quality; ///< Quality to map the rest of the values to.
173 /* This is overloaded to be both kbps per channel in ABR mode, and
174 * requested quality in constant quality mode.
175 */
176 float st_lrm; ///< short threshold for L, R, and M channels
177 } PsyLamePreset;
178
179 /**
180 * LAME psy model preset table for ABR
181 */
182 static const PsyLamePreset psy_abr_map[] = {
183 /* TODO: Tuning. These were taken from LAME. */
184 /* kbps/ch st_lrm */
185 { 8, 7.60},
186 { 16, 7.60},
187 { 24, 7.60},
188 { 32, 7.60},
189 { 40, 7.60},
190 { 48, 7.60},
191 { 56, 7.60},
192 { 64, 7.40},
193 { 80, 7.00},
194 { 96, 6.60},
195 {112, 6.20},
196 {128, 6.20},
197 {160, 6.20}
198 };
199
200 /**
201 * LAME psy model preset table for constant quality
202 */
203 static const PsyLamePreset psy_vbr_map[] = {
204 /* vbr_q st_lrm */
205 { 0, 4.20},
206 { 1, 4.20},
207 { 2, 4.20},
208 { 3, 4.20},
209 { 4, 4.20},
210 { 5, 4.20},
211 { 6, 4.20},
212 { 7, 4.20},
213 { 8, 4.20},
214 { 9, 4.20},
215 {10, 4.20}
216 };
217
218 /**
219 * LAME psy model FIR coefficient table
220 */
221 static const float psy_fir_coeffs[] = {
222 -8.65163e-18 * 2, -0.00851586 * 2, -6.74764e-18 * 2, 0.0209036 * 2,
223 -3.36639e-17 * 2, -0.0438162 * 2, -1.54175e-17 * 2, 0.0931738 * 2,
224 -5.52212e-17 * 2, -0.313819 * 2
225 };
226
227 /**
228 * Calculate the ABR attack threshold from the above LAME psymodel table.
229 */
230 23 static float lame_calc_attack_threshold(int bitrate)
231 {
232 /* Assume max bitrate to start with */
233 23 int lower_range = 12, upper_range = 12;
234 23 int lower_range_kbps = psy_abr_map[12].quality;
235 23 int upper_range_kbps = psy_abr_map[12].quality;
236 int i;
237
238 /* Determine which bitrates the value specified falls between.
239 * If the loop ends without breaking our above assumption of 320kbps was correct.
240 */
241
2/2
✓ Branch 0 taken 194 times.
✓ Branch 1 taken 4 times.
 198 for (i = 1; i < 13; i++) {
242
2/2
✓ Branch 0 taken 19 times.
✓ Branch 1 taken 175 times.
 194 if (FFMAX(bitrate, psy_abr_map[i].quality) != bitrate) {
243 19 upper_range = i;
244 19 upper_range_kbps = psy_abr_map[i ].quality;
245 19 lower_range = i - 1;
246 19 lower_range_kbps = psy_abr_map[i - 1].quality;
247 19 break; /* Upper range found */
248 }
249 }
250
251 /* Determine which range the value specified is closer to */
252
2/2
✓ Branch 0 taken 19 times.
✓ Branch 1 taken 4 times.
 23 if ((upper_range_kbps - bitrate) > (bitrate - lower_range_kbps))
253 19 return psy_abr_map[lower_range].st_lrm;
254 4 return psy_abr_map[upper_range].st_lrm;
255 }
256
257 /**
258 * LAME psy model specific initialization
259 */
260 10 static av_cold void lame_window_init(AacPsyContext *ctx, AVCodecContext *avctx)
261 {
262 int i, j;
263
264
2/2
✓ Branch 0 taken 23 times.
✓ Branch 1 taken 10 times.
 33 for (i = 0; i < avctx->ch_layout.nb_channels; i++) {
265 23 AacPsyChannel *pch = &ctx->ch[i];
266
267
1/2
✗ Branch 0 not taken.
✓ Branch 1 taken 23 times.
 23 if (avctx->flags & AV_CODEC_FLAG_QSCALE)
268 ✗ pch->attack_threshold = psy_vbr_map[av_clip(avctx->global_quality / FF_QP2LAMBDA, 0, 10)].st_lrm;
269 else
270 23 pch->attack_threshold = lame_calc_attack_threshold(avctx->bit_rate / avctx->ch_layout.nb_channels / 1000);
271
272
2/2
✓ Branch 0 taken 368 times.
✓ Branch 1 taken 23 times.
 391 for (j = 0; j < AAC_NUM_BLOCKS_SHORT * PSY_LAME_NUM_SUBBLOCKS; j++)
273 368 pch->prev_energy_subshort[j] = 10.0f;
274 }
275 10 }
276
277 /**
278 * Calculate Bark value for given line.
279 */
280 640 static av_cold float calc_bark(float f)
281 {
282 640 return 13.3f * atanf(0.00076f * f) + 3.5f * atanf((f / 7500.0f) * (f / 7500.0f));
283 }
284
285 #define ATH_ADD 4
286 /**
287 * Calculate ATH value for given frequency.
288 * Borrowed from Lame.
289 */
290 13036 static av_cold float ath(float f, float add)
291 {
292 13036 f /= 1000.0f;
293 13036 return 3.64 * pow(f, -0.8)
294 13036 - 6.8 * exp(-0.6 * (f - 3.4) * (f - 3.4))
295 13036 + 6.0 * exp(-0.15 * (f - 8.7) * (f - 8.7))
296 13036 + (0.6 + 0.04 * add) * 0.001 * f * f * f * f;
297 }
298
299 10 static av_cold int psy_3gpp_init(FFPsyContext *ctx) {
300 AacPsyContext *pctx;
301 float bark;
302 int i, j, g, start;
303 float prev, minscale, minath, minsnr, pe_min;
304
1/2
✓ Branch 0 taken 10 times.
✗ Branch 1 not taken.
 10 int chan_bitrate = ctx->avctx->bit_rate / ((ctx->avctx->flags & AV_CODEC_FLAG_QSCALE) ? 2.0f : ctx->avctx->ch_layout.nb_channels);
305
306
5/8
✓ Branch 0 taken 5 times.
✓ Branch 1 taken 5 times.
✗ Branch 2 not taken.
✓ Branch 3 taken 5 times.
✓ Branch 4 taken 5 times.
✗ Branch 5 not taken.
✗ Branch 6 not taken.
✓ Branch 7 taken 5 times.
 10 const int bandwidth = ctx->cutoff ? ctx->cutoff : AAC_CUTOFF(ctx->avctx);
307 10 const float num_bark = calc_bark((float)bandwidth);
308
309
1/2
✗ Branch 0 not taken.
✓ Branch 1 taken 10 times.
 10 if (bandwidth <= 0)
310 ✗ return AVERROR(EINVAL);
311
312 10 ctx->model_priv_data = av_mallocz(sizeof(AacPsyContext));
313
1/2
✗ Branch 0 not taken.
✓ Branch 1 taken 10 times.
 10 if (!ctx->model_priv_data)
314 ✗ return AVERROR(ENOMEM);
315 10 pctx = ctx->model_priv_data;
316
1/2
✗ Branch 0 not taken.
✓ Branch 1 taken 10 times.
 10 pctx->global_quality = (ctx->avctx->global_quality ? ctx->avctx->global_quality : 120) * 0.01f;
317
318
1/2
✗ Branch 0 not taken.
✓ Branch 1 taken 10 times.
 10 if (ctx->avctx->flags & AV_CODEC_FLAG_QSCALE) {
319 /* Use the target average bitrate to compute spread parameters */
320 ✗ chan_bitrate = (int)(chan_bitrate / 120.0 * (ctx->avctx->global_quality ? ctx->avctx->global_quality : 120));
321 }
322
323 10 pctx->chan_bitrate = chan_bitrate;
324 10 pctx->frame_bits = FFMIN(2560, chan_bitrate * AAC_BLOCK_SIZE_LONG / ctx->avctx->sample_rate);
325 10 pctx->pe.min = 8.0f * AAC_BLOCK_SIZE_LONG * bandwidth / (ctx->avctx->sample_rate * 2.0f);
326 10 pctx->pe.max = 12.0f * AAC_BLOCK_SIZE_LONG * bandwidth / (ctx->avctx->sample_rate * 2.0f);
327 10 ctx->bitres.size = 6144 - pctx->frame_bits;
328 10 ctx->bitres.size -= ctx->bitres.size % 8;
329 10 pctx->fill_level = ctx->bitres.size;
330 10 minath = ath(3410 - 0.733 * ATH_ADD, ATH_ADD);
331
2/2
✓ Branch 0 taken 20 times.
✓ Branch 1 taken 10 times.
 30 for (j = 0; j < 2; j++) {
332 20 AacPsyCoeffs *coeffs = pctx->psy_coef[j];
333 20 const uint8_t *band_sizes = ctx->bands[j];
334
2/2
✓ Branch 0 taken 10 times.
✓ Branch 1 taken 10 times.
 20 float line_to_frequency = ctx->avctx->sample_rate / (j ? 256.f : 2048.0f);
335
2/2
✓ Branch 0 taken 10 times.
✓ Branch 1 taken 10 times.
 20 float avg_chan_bits = chan_bitrate * (j ? 128.0f : 1024.0f) / ctx->avctx->sample_rate;
336 /* reference encoder uses 2.4% here instead of 60% like the spec says */
337 20 float bark_pe = 0.024f * PSY_3GPP_BITS_TO_PE(avg_chan_bits) / num_bark;
338
2/2
✓ Branch 0 taken 10 times.
✓ Branch 1 taken 10 times.
 20 float en_spread_low = j ? PSY_3GPP_EN_SPREAD_LOW_S : PSY_3GPP_EN_SPREAD_LOW_L;
339 /* High energy spreading for long blocks <= 22kbps/channel and short blocks are the same. */
340
3/4
✓ Branch 0 taken 10 times.
✓ Branch 1 taken 10 times.
✗ Branch 2 not taken.
✓ Branch 3 taken 10 times.
 20 float en_spread_hi = (j || (chan_bitrate <= 22.0f)) ? PSY_3GPP_EN_SPREAD_HI_S : PSY_3GPP_EN_SPREAD_HI_L1;
341
342 20 i = 0;
343 20 prev = 0.0;
344
2/2
✓ Branch 0 taken 630 times.
✓ Branch 1 taken 20 times.
 650 for (g = 0; g < ctx->num_bands[j]; g++) {
345 630 i += band_sizes[g];
346 630 bark = calc_bark((i-1) * line_to_frequency);
347 630 coeffs[g].barks = (bark + prev) / 2.0;
348 630 prev = bark;
349 }
350
2/2
✓ Branch 0 taken 610 times.
✓ Branch 1 taken 20 times.
 630 for (g = 0; g < ctx->num_bands[j] - 1; g++) {
351 610 AacPsyCoeffs *coeff = &coeffs[g];
352 610 float bark_width = coeffs[g+1].barks - coeffs->barks;
353 610 coeff->spread_low[0] = ff_exp10(-bark_width * PSY_3GPP_THR_SPREAD_LOW);
354 610 coeff->spread_hi [0] = ff_exp10(-bark_width * PSY_3GPP_THR_SPREAD_HI);
355 610 coeff->spread_low[1] = ff_exp10(-bark_width * en_spread_low);
356 610 coeff->spread_hi [1] = ff_exp10(-bark_width * en_spread_hi);
357 610 pe_min = bark_pe * bark_width;
358 610 minsnr = exp2(pe_min / band_sizes[g]) - 1.5f;
359 610 coeff->min_snr = av_clipf(1.0f / minsnr, PSY_SNR_25DB, PSY_SNR_1DB);
360 }
361 20 start = 0;
362
2/2
✓ Branch 0 taken 630 times.
✓ Branch 1 taken 20 times.
 650 for (g = 0; g < ctx->num_bands[j]; g++) {
363 630 minscale = ath(start * line_to_frequency, ATH_ADD);
364
2/2
✓ Branch 0 taken 10890 times.
✓ Branch 1 taken 630 times.
 11520 for (i = 1; i < band_sizes[g]; i++)
365
2/2
✓ Branch 1 taken 1506 times.
✓ Branch 2 taken 9384 times.
 10890 minscale = FFMIN(minscale, ath((start + i) * line_to_frequency, ATH_ADD));
366 630 coeffs[g].ath = minscale - minath;
367 630 start += band_sizes[g];
368 }
369 }
370
371 10 pctx->ch = av_calloc(ctx->avctx->ch_layout.nb_channels, sizeof(*pctx->ch));
372
1/2
✗ Branch 0 not taken.
✓ Branch 1 taken 10 times.
 10 if (!pctx->ch) {
373 ✗ av_freep(&ctx->model_priv_data);
374 ✗ return AVERROR(ENOMEM);
375 }
376
377 10 lame_window_init(pctx, ctx->avctx);
378
379 10 return 0;
380 }
381
382 /**
383 * IIR filter used in block switching decision
384 */
385 ✗ static float iir_filter(int in, float state[2])
386 {
387 float ret;
388
389 ✗ ret = 0.7548f * (in - state[0]) + 0.5095f * state[1];
390 ✗ state[0] = in;
391 ✗ state[1] = ret;
392 ✗ return ret;
393 }
394
395 /**
396 * window grouping information stored as bits (0 - new group, 1 - group continues)
397 */
398 static const uint8_t window_grouping[9] = {
399 0xB6, 0x6C, 0xD8, 0xB2, 0x66, 0xC6, 0x96, 0x36, 0x36
400 };
401
402 /**
403 * Tell encoder which window types to use.
404 * @see 3GPP TS26.403 5.4.1 "Blockswitching"
405 */
406 ✗ av_unused static FFPsyWindowInfo psy_3gpp_window(FFPsyContext *ctx,
407 const int16_t *audio,
408 const int16_t *la,
409 int channel, int prev_type)
410 {
411 int i, j;
412 ✗ int br = ((AacPsyContext*)ctx->model_priv_data)->chan_bitrate;
413 ✗ int attack_ratio = br <= 16000 ? 18 : 10;
414 ✗ AacPsyContext *pctx = (AacPsyContext*) ctx->model_priv_data;
415 ✗ AacPsyChannel *pch = &pctx->ch[channel];
416 ✗ uint8_t grouping = 0;
417 ✗ int next_type = pch->next_window_seq;
418 ✗ FFPsyWindowInfo wi = { { 0 } };
419
420 ✗ if (la) {
421 float s[8], v;
422 ✗ int switch_to_eight = 0;
423 ✗ float sum = 0.0, sum2 = 0.0;
424 ✗ int attack_n = 0;
425 ✗ int stay_short = 0;
426 ✗ for (i = 0; i < 8; i++) {
427 ✗ for (j = 0; j < 128; j++) {
428 ✗ v = iir_filter(la[i*128+j], pch->iir_state);
429 ✗ sum += v*v;
430 }
431 ✗ s[i] = sum;
432 ✗ sum2 += sum;
433 }
434 ✗ for (i = 0; i < 8; i++) {
435 ✗ if (s[i] > pch->win_energy * attack_ratio) {
436 ✗ attack_n = i + 1;
437 ✗ switch_to_eight = 1;
438 ✗ break;
439 }
440 }
441 ✗ pch->win_energy = pch->win_energy*7/8 + sum2/64;
442
443 ✗ wi.window_type[1] = prev_type;
444 ✗ switch (prev_type) {
445 ✗ case ONLY_LONG_SEQUENCE:
446 ✗ wi.window_type[0] = switch_to_eight ? LONG_START_SEQUENCE : ONLY_LONG_SEQUENCE;
447 ✗ next_type = switch_to_eight ? EIGHT_SHORT_SEQUENCE : ONLY_LONG_SEQUENCE;
448 ✗ break;
449 ✗ case LONG_START_SEQUENCE:
450 ✗ wi.window_type[0] = EIGHT_SHORT_SEQUENCE;
451 ✗ grouping = pch->next_grouping;
452 ✗ next_type = switch_to_eight ? EIGHT_SHORT_SEQUENCE : LONG_STOP_SEQUENCE;
453 ✗ break;
454 ✗ case LONG_STOP_SEQUENCE:
455 ✗ wi.window_type[0] = switch_to_eight ? LONG_START_SEQUENCE : ONLY_LONG_SEQUENCE;
456 ✗ next_type = switch_to_eight ? EIGHT_SHORT_SEQUENCE : ONLY_LONG_SEQUENCE;
457 ✗ break;
458 ✗ case EIGHT_SHORT_SEQUENCE:
459 ✗ stay_short = next_type == EIGHT_SHORT_SEQUENCE || switch_to_eight;
460 ✗ wi.window_type[0] = stay_short ? EIGHT_SHORT_SEQUENCE : LONG_STOP_SEQUENCE;
461 ✗ grouping = next_type == EIGHT_SHORT_SEQUENCE ? pch->next_grouping : 0;
462 ✗ next_type = switch_to_eight ? EIGHT_SHORT_SEQUENCE : LONG_STOP_SEQUENCE;
463 ✗ break;
464 }
465
466 ✗ pch->next_grouping = window_grouping[attack_n];
467 ✗ pch->next_window_seq = next_type;
468 } else {
469 ✗ for (i = 0; i < 3; i++)
470 ✗ wi.window_type[i] = prev_type;
471 ✗ grouping = (prev_type == EIGHT_SHORT_SEQUENCE) ? window_grouping[0] : 0;
472 }
473
474 ✗ wi.window_shape = 1;
475 ✗ if (wi.window_type[0] != EIGHT_SHORT_SEQUENCE) {
476 ✗ wi.num_windows = 1;
477 ✗ wi.grouping[0] = 1;
478 } else {
479 ✗ int lastgrp = 0;
480 ✗ wi.num_windows = 8;
481 ✗ for (i = 0; i < 8; i++) {
482 ✗ if (!((grouping >> i) & 1))
483 ✗ lastgrp = i;
484 ✗ wi.grouping[lastgrp]++;
485 }
486 }
487
488 ✗ return wi;
489 }
490
491 /* 5.6.1.2 "Calculation of Bit Demand" */
492 9375 static int calc_bit_demand(AacPsyContext *ctx, float pe, int bits, int size,
493 int short_window)
494 {
495
2/2
✓ Branch 0 taken 197 times.
✓ Branch 1 taken 9178 times.
 9375 const float bitsave_slope = short_window ? PSY_3GPP_SAVE_SLOPE_S : PSY_3GPP_SAVE_SLOPE_L;
496
2/2
✓ Branch 0 taken 197 times.
✓ Branch 1 taken 9178 times.
 9375 const float bitsave_add = short_window ? PSY_3GPP_SAVE_ADD_S : PSY_3GPP_SAVE_ADD_L;
497
2/2
✓ Branch 0 taken 197 times.
✓ Branch 1 taken 9178 times.
 9375 const float bitspend_slope = short_window ? PSY_3GPP_SPEND_SLOPE_S : PSY_3GPP_SPEND_SLOPE_L;
498
2/2
✓ Branch 0 taken 197 times.
✓ Branch 1 taken 9178 times.
 9375 const float bitspend_add = short_window ? PSY_3GPP_SPEND_ADD_S : PSY_3GPP_SPEND_ADD_L;
499 9375 const float clip_low = short_window ? PSY_3GPP_CLIP_LO_S : PSY_3GPP_CLIP_LO_L;
500
2/2
✓ Branch 0 taken 197 times.
✓ Branch 1 taken 9178 times.
 9375 const float clip_high = short_window ? PSY_3GPP_CLIP_HI_S : PSY_3GPP_CLIP_HI_L;
501 float clipped_pe, bit_save, bit_spend, bit_factor, fill_level, forgetful_min_pe;
502
503 9375 ctx->fill_level += ctx->frame_bits - bits;
504 9375 ctx->fill_level = av_clip(ctx->fill_level, 0, size);
505 9375 fill_level = av_clipf((float)ctx->fill_level / size, clip_low, clip_high);
506 9375 clipped_pe = av_clipf(pe, ctx->pe.min, ctx->pe.max);
507 9375 bit_save = (fill_level + bitsave_add) * bitsave_slope;
508 assert(bit_save <= 0.3f && bit_save >= -0.05000001f);
509 9375 bit_spend = (fill_level + bitspend_add) * bitspend_slope;
510 assert(bit_spend <= 0.5f && bit_spend >= -0.1f);
511 /* The bit factor graph in the spec is obviously incorrect.
512 * bit_spend + ((bit_spend - bit_spend))...
513 * The reference encoder subtracts everything from 1, but also seems incorrect.
514 * 1 - bit_save + ((bit_spend + bit_save))...
515 * Hopefully below is correct.
516 */
517 9375 bit_factor = 1.0f - bit_save + ((bit_spend - bit_save) / (ctx->pe.max - ctx->pe.min)) * (clipped_pe - ctx->pe.min);
518 /* NOTE: The reference encoder attempts to center pe max/min around the current pe.
519 * Here we do that by slowly forgetting pe.min when pe stays in a range that makes
520 * it unlikely (ie: above the mean)
521 */
522
2/2
✓ Branch 0 taken 19 times.
✓ Branch 1 taken 9356 times.
 9375 ctx->pe.max = FFMAX(pe, ctx->pe.max);
523 18750 forgetful_min_pe = ((ctx->pe.min * PSY_PE_FORGET_SLOPE)
524
2/2
✓ Branch 0 taken 2769 times.
✓ Branch 1 taken 6606 times.
 9375 + FFMAX(ctx->pe.min, pe * (pe / ctx->pe.max))) / (PSY_PE_FORGET_SLOPE + 1);
525
2/2
✓ Branch 0 taken 9297 times.
✓ Branch 1 taken 78 times.
 9375 ctx->pe.min = FFMIN(pe, forgetful_min_pe);
526
527 /* NOTE: allocate a minimum of 1/8th average frame bits, to avoid
528 * reservoir starvation from producing zero-bit frames
529 */
530
6/6
✓ Branch 0 taken 9361 times.
✓ Branch 1 taken 14 times.
✓ Branch 2 taken 1147 times.
✓ Branch 3 taken 8228 times.
✓ Branch 4 taken 1133 times.
✓ Branch 5 taken 14 times.
 9375 return FFMIN(
531 ctx->frame_bits * bit_factor,
532 FFMAX(ctx->frame_bits + size - bits, ctx->frame_bits / 8));
533 }
534
535 1562498 static float calc_pe_3gpp(AacPsyBand *band)
536 {
537 float pe, a;
538
539 1562498 band->pe = 0.0f;
540 1562498 band->pe_const = 0.0f;
541 1562498 band->active_lines = 0.0f;
542
2/2
✓ Branch 0 taken 1517484 times.
✓ Branch 1 taken 45014 times.
 1562498 if (band->energy > band->thr) {
543 1517484 a = log2f(band->energy);
544 1517484 pe = a - log2f(band->thr);
545 1517484 band->active_lines = band->nz_lines;
546
2/2
✓ Branch 0 taken 730962 times.
✓ Branch 1 taken 786522 times.
 1517484 if (pe < PSY_3GPP_C1) {
547 730962 pe = pe * PSY_3GPP_C3 + PSY_3GPP_C2;
548 730962 a = a * PSY_3GPP_C3 + PSY_3GPP_C2;
549 730962 band->active_lines *= PSY_3GPP_C3;
550 }
551 1517484 band->pe = pe * band->nz_lines;
552 1517484 band->pe_const = a * band->nz_lines;
553 }
554
555 1562498 return band->pe;
556 }
557
558 22813 static float calc_reduction_3gpp(float a, float desired_pe, float pe,
559 float active_lines)
560 {
561 float thr_avg, reduction;
562
563
2/2
✓ Branch 0 taken 24 times.
✓ Branch 1 taken 22789 times.
 22813 if(active_lines == 0.0)
564 24 return 0;
565
566 22789 thr_avg = exp2f((a - pe) / (4.0f * active_lines));
567 22789 reduction = exp2f((a - desired_pe) / (4.0f * active_lines)) - thr_avg;
568
569
2/2
✓ Branch 0 taken 20760 times.
✓ Branch 1 taken 2029 times.
 22789 return FFMAX(reduction, 0.0f);
570 }
571
572 1090712 static float calc_reduced_thr_3gpp(AacPsyBand *band, float min_snr,
573 float reduction)
574 {
575 1090712 float thr = band->thr;
576
577
2/2
✓ Branch 0 taken 1064622 times.
✓ Branch 1 taken 26090 times.
 1090712 if (band->energy > thr) {
578 1064622 thr = sqrtf(thr);
579 1064622 thr = sqrtf(thr) + reduction;
580 1064622 thr *= thr;
581 1064622 thr *= thr;
582
583 /* This deviates from the 3GPP spec to match the reference encoder.
584 * It performs min(thr_reduced, max(thr, energy/min_snr)) only for bands
585 * that have hole avoidance on (active or inactive). It always reduces the
586 * threshold of bands with hole avoidance off.
587 */
588
4/4
✓ Branch 0 taken 335892 times.
✓ Branch 1 taken 728730 times.
✓ Branch 2 taken 335824 times.
✓ Branch 3 taken 68 times.
 1064622 if (thr > band->energy * min_snr && band->avoid_holes != PSY_3GPP_AH_NONE) {
589
2/2
✓ Branch 0 taken 4668 times.
✓ Branch 1 taken 331156 times.
 335824 thr = FFMAX(band->thr, band->energy * min_snr);
590 335824 band->avoid_holes = PSY_3GPP_AH_ACTIVE;
591 }
592 }
593
594 1090712 return thr;
595 }
596
597 9375 static void calc_thr_3gpp(const FFPsyWindowInfo *wi, const int num_bands, AacPsyChannel *pch,
598 const uint8_t *band_sizes, const float *coefs, const int cutoff)
599 {
600 int i, w, g;
601 9375 int start = 0, wstart = 0;
602
2/2
✓ Branch 0 taken 10754 times.
✓ Branch 1 taken 9375 times.
 20129 for (w = 0; w < wi->num_windows*16; w += 16) {
603 10754 wstart = 0;
604
2/2
✓ Branch 0 taken 471786 times.
✓ Branch 1 taken 10754 times.
 482540 for (g = 0; g < num_bands; g++) {
605 471786 AacPsyBand *band = &pch->band[w+g];
606
607 471786 float form_factor = 0.0f;
608 float Temp;
609 471786 band->energy = 0.0f;
610
2/2
✓ Branch 0 taken 464322 times.
✓ Branch 1 taken 7464 times.
 471786 if (wstart < cutoff) {
611
2/2
✓ Branch 0 taken 9293760 times.
✓ Branch 1 taken 464322 times.
 9758082 for (i = 0; i < band_sizes[g]; i++) {
612 9293760 band->energy += coefs[start+i] * coefs[start+i];
613 9293760 form_factor += sqrtf(fabs(coefs[start+i]));
614 }
615 }
616
2/2
✓ Branch 0 taken 455509 times.
✓ Branch 1 taken 16277 times.
 471786 Temp = band->energy > 0 ? sqrtf((float)band_sizes[g] / band->energy) : 0;
617 471786 band->thr = band->energy * 0.001258925f;
618 471786 band->nz_lines = form_factor * sqrtf(Temp);
619
620 471786 start += band_sizes[g];
621 471786 wstart += band_sizes[g];
622 }
623 }
624 9375 }
625
626 6370 static void psy_hp_filter(const float *firbuf, float *hpfsmpl, const float *psy_fir_coeffs)
627 {
628 int i, j;
629
2/2
✓ Branch 0 taken 6522880 times.
✓ Branch 1 taken 6370 times.
 6529250 for (i = 0; i < AAC_BLOCK_SIZE_LONG; i++) {
630 float sum1, sum2;
631 6522880 sum1 = firbuf[i + (PSY_LAME_FIR_LEN - 1) / 2];
632 6522880 sum2 = 0.0;
633
2/2
✓ Branch 0 taken 32614400 times.
✓ Branch 1 taken 6522880 times.
 39137280 for (j = 0; j < ((PSY_LAME_FIR_LEN - 1) / 2) - 1; j += 2) {
634 32614400 sum1 += psy_fir_coeffs[j] * (firbuf[i + j] + firbuf[i + PSY_LAME_FIR_LEN - j]);
635 32614400 sum2 += psy_fir_coeffs[j + 1] * (firbuf[i + j + 1] + firbuf[i + PSY_LAME_FIR_LEN - j - 1]);
636 }
637 /* NOTE: The LAME psymodel expects it's input in the range -32768 to 32768.
638 * Tuning this for normalized floats would be difficult. */
639 6522880 hpfsmpl[i] = (sum1 + sum2) * 32768.0f;
640 }
641 6370 }
642
643 /**
644 * Calculate band thresholds as suggested in 3GPP TS26.403
645 */
646 9375 static void psy_3gpp_analyze_channel(FFPsyContext *ctx, int channel,
647 const float *coefs, const FFPsyWindowInfo *wi)
648 {
649 9375 AacPsyContext *pctx = (AacPsyContext*) ctx->model_priv_data;
650 9375 AacPsyChannel *pch = &pctx->ch[channel];
651 int i, w, g;
652 9375 float desired_bits, desired_pe, delta_pe, reduction= NAN, spread_en[128] = {0};
653 9375 float a = 0.0f, active_lines = 0.0f, norm_fac = 0.0f;
654
1/4
✗ Branch 0 not taken.
✓ Branch 1 taken 9375 times.
✗ Branch 2 not taken.
✗ Branch 3 not taken.
 9375 float pe = pctx->chan_bitrate > 32000 ? 0.0f : FFMAX(50.0f, 100.0f - pctx->chan_bitrate * 100.0f / 32000.0f);
655
2/2
✓ Branch 0 taken 197 times.
✓ Branch 1 taken 9178 times.
 9375 const int num_bands = ctx->num_bands[wi->num_windows == 8];
656
2/2
✓ Branch 0 taken 197 times.
✓ Branch 1 taken 9178 times.
 9375 const uint8_t *band_sizes = ctx->bands[wi->num_windows == 8];
657 9375 AacPsyCoeffs *coeffs = pctx->psy_coef[wi->num_windows == 8];
658
2/2
✓ Branch 0 taken 197 times.
✓ Branch 1 taken 9178 times.
 9375 const float avoid_hole_thr = wi->num_windows == 8 ? PSY_3GPP_AH_THR_SHORT : PSY_3GPP_AH_THR_LONG;
659
5/8
✓ Branch 0 taken 4900 times.
✓ Branch 1 taken 4475 times.
✗ Branch 2 not taken.
✓ Branch 3 taken 4475 times.
✓ Branch 4 taken 4475 times.
✗ Branch 5 not taken.
✗ Branch 6 not taken.
✓ Branch 7 taken 4475 times.
 9375 const int bandwidth = ctx->cutoff ? ctx->cutoff : AAC_CUTOFF(ctx->avctx);
660 9375 const int cutoff = bandwidth * 2048 / wi->num_windows / ctx->avctx->sample_rate;
661
662 //calculate energies, initial thresholds and related values - 5.4.2 "Threshold Calculation"
663 9375 calc_thr_3gpp(wi, num_bands, pch, band_sizes, coefs, cutoff);
664
665 //modify thresholds and energies - spread, threshold in quiet, pre-echo control
666
2/2
✓ Branch 0 taken 10754 times.
✓ Branch 1 taken 9375 times.
 20129 for (w = 0; w < wi->num_windows*16; w += 16) {
667 10754 AacPsyBand *bands = &pch->band[w];
668
669 /* 5.4.2.3 "Spreading" & 5.4.3 "Spread Energy Calculation" */
670 10754 spread_en[0] = bands[0].energy;
671
2/2
✓ Branch 0 taken 461032 times.
✓ Branch 1 taken 10754 times.
 471786 for (g = 1; g < num_bands; g++) {
672
2/2
✓ Branch 0 taken 444776 times.
✓ Branch 1 taken 16256 times.
 461032 bands[g].thr = FFMAX(bands[g].thr, bands[g-1].thr * coeffs[g].spread_hi[0]);
673
2/2
✓ Branch 0 taken 445266 times.
✓ Branch 1 taken 15766 times.
 461032 spread_en[w+g] = FFMAX(bands[g].energy, spread_en[w+g-1] * coeffs[g].spread_hi[1]);
674 }
675
2/2
✓ Branch 0 taken 461032 times.
✓ Branch 1 taken 10754 times.
 471786 for (g = num_bands - 2; g >= 0; g--) {
676
2/2
✓ Branch 0 taken 447815 times.
✓ Branch 1 taken 13217 times.
 461032 bands[g].thr = FFMAX(bands[g].thr, bands[g+1].thr * coeffs[g].spread_low[0]);
677
2/2
✓ Branch 0 taken 445857 times.
✓ Branch 1 taken 15175 times.
 461032 spread_en[w+g] = FFMAX(spread_en[w+g], spread_en[w+g+1] * coeffs[g].spread_low[1]);
678 }
679 //5.4.2.4 "Threshold in quiet"
680
2/2
✓ Branch 0 taken 471786 times.
✓ Branch 1 taken 10754 times.
 482540 for (g = 0; g < num_bands; g++) {
681 471786 AacPsyBand *band = &bands[g];
682
683
2/2
✓ Branch 0 taken 345402 times.
✓ Branch 1 taken 126384 times.
 471786 band->thr_quiet = band->thr = FFMAX(band->thr, coeffs[g].ath);
684 //5.4.2.5 "Pre-echo control"
685
6/6
✓ Branch 0 taken 467621 times.
✓ Branch 1 taken 4165 times.
✓ Branch 2 taken 448315 times.
✓ Branch 3 taken 19306 times.
✓ Branch 4 taken 447041 times.
✓ Branch 5 taken 1274 times.
 471786 if (!(wi->window_type[0] == LONG_STOP_SEQUENCE || (!w && wi->window_type[1] == LONG_START_SEQUENCE)))
686
6/6
✓ Branch 0 taken 53844 times.
✓ Branch 1 taken 412503 times.
✓ Branch 2 taken 4450 times.
✓ Branch 3 taken 461897 times.
✓ Branch 4 taken 49394 times.
✓ Branch 5 taken 412503 times.
 466347 band->thr = FFMAX(PSY_3GPP_RPEMIN*band->thr, FFMIN(band->thr,
687 PSY_3GPP_RPELEV*pch->prev_band[w+g].thr_quiet));
688
689 /* 5.6.1.3.1 "Preparatory steps of the perceptual entropy calculation" */
690 471786 pe += calc_pe_3gpp(band);
691 471786 a += band->pe_const;
692 471786 active_lines += band->active_lines;
693
694 /* 5.6.1.3.3 "Selection of the bands for avoidance of holes" */
695
3/4
✓ Branch 0 taken 471591 times.
✓ Branch 1 taken 195 times.
✗ Branch 2 not taken.
✓ Branch 3 taken 471591 times.
 471786 if (spread_en[w+g] * avoid_hole_thr > band->energy || coeffs[g].min_snr > 1.0f)
696 195 band->avoid_holes = PSY_3GPP_AH_NONE;
697 else
698 471591 band->avoid_holes = PSY_3GPP_AH_INACTIVE;
699 }
700 }
701
702 /* 5.6.1.3.2 "Calculation of the desired perceptual entropy" */
703 9375 ctx->ch[channel].entropy = pe;
704
1/2
✗ Branch 0 not taken.
✓ Branch 1 taken 9375 times.
 9375 if (ctx->avctx->flags & AV_CODEC_FLAG_QSCALE) {
705 /* (2.5 * 120) achieves almost transparent rate, and we want to give
706 * ample room downwards, so we make that equivalent to QSCALE=2.4
707 */
708 ✗ desired_pe = pe * (ctx->avctx->global_quality ? ctx->avctx->global_quality : 120) / (2 * 2.5f * 120.0f);
709 ✗ desired_bits = FFMIN(2560, PSY_3GPP_PE_TO_BITS(desired_pe));
710 ✗ desired_pe = PSY_3GPP_BITS_TO_PE(desired_bits); // reflect clipping
711
712 /* PE slope smoothing */
713 ✗ if (ctx->bitres.bits > 0) {
714 ✗ desired_bits = FFMIN(2560, PSY_3GPP_PE_TO_BITS(desired_pe));
715 ✗ desired_pe = PSY_3GPP_BITS_TO_PE(desired_bits); // reflect clipping
716 }
717
718 ✗ pctx->pe.max = FFMAX(pe, pctx->pe.max);
719 ✗ pctx->pe.min = FFMIN(pe, pctx->pe.min);
720 } else {
721 9375 desired_bits = calc_bit_demand(pctx, pe, ctx->bitres.bits, ctx->bitres.size, wi->num_windows == 8);
722 9375 desired_pe = PSY_3GPP_BITS_TO_PE(desired_bits);
723
724 /* NOTE: PE correction is kept simple. During initial testing it had very
725 * little effect on the final bitrate. Probably a good idea to come
726 * back and do more testing later.
727 */
728
2/2
✓ Branch 0 taken 9314 times.
✓ Branch 1 taken 61 times.
 9375 if (ctx->bitres.bits > 0)
729 9314 desired_pe *= av_clipf(pctx->pe.previous / PSY_3GPP_BITS_TO_PE(ctx->bitres.bits),
730 0.85f, 1.15f);
731 }
732 9375 pctx->pe.previous = PSY_3GPP_BITS_TO_PE(desired_bits);
733 9375 ctx->bitres.alloc = desired_bits;
734
735
2/2
✓ Branch 0 taken 9272 times.
✓ Branch 1 taken 103 times.
 9375 if (desired_pe < pe) {
736 /* 5.6.1.3.4 "First Estimation of the reduction value" */
737
2/2
✓ Branch 0 taken 10147 times.
✓ Branch 1 taken 9272 times.
 19419 for (w = 0; w < wi->num_windows*16; w += 16) {
738 10147 reduction = calc_reduction_3gpp(a, desired_pe, pe, active_lines);
739 10147 pe = 0.0f;
740 10147 a = 0.0f;
741 10147 active_lines = 0.0f;
742
2/2
✓ Branch 0 taken 462203 times.
✓ Branch 1 taken 10147 times.
 472350 for (g = 0; g < num_bands; g++) {
743 462203 AacPsyBand *band = &pch->band[w+g];
744
745 462203 band->thr = calc_reduced_thr_3gpp(band, coeffs[g].min_snr, reduction);
746 /* recalculate PE */
747 462203 pe += calc_pe_3gpp(band);
748 462203 a += band->pe_const;
749 462203 active_lines += band->active_lines;
750 }
751 }
752
753 /* 5.6.1.3.5 "Second Estimation of the reduction value" */
754
2/2
✓ Branch 0 taken 12666 times.
✓ Branch 1 taken 3394 times.
 16060 for (i = 0; i < 2; i++) {
755 12666 float pe_no_ah = 0.0f, desired_pe_no_ah;
756 12666 active_lines = a = 0.0f;
757
2/2
✓ Branch 0 taken 13541 times.
✓ Branch 1 taken 12666 times.
 26207 for (w = 0; w < wi->num_windows*16; w += 16) {
758
2/2
✓ Branch 0 taken 628509 times.
✓ Branch 1 taken 13541 times.
 642050 for (g = 0; g < num_bands; g++) {
759 628509 AacPsyBand *band = &pch->band[w+g];
760
761
2/2
✓ Branch 0 taken 501873 times.
✓ Branch 1 taken 126636 times.
 628509 if (band->avoid_holes != PSY_3GPP_AH_ACTIVE) {
762 501873 pe_no_ah += band->pe;
763 501873 a += band->pe_const;
764 501873 active_lines += band->active_lines;
765 }
766 }
767 }
768
2/2
✓ Branch 0 taken 12629 times.
✓ Branch 1 taken 37 times.
 12666 desired_pe_no_ah = FFMAX(desired_pe - (pe - pe_no_ah), 0.0f);
769
1/2
✓ Branch 0 taken 12666 times.
✗ Branch 1 not taken.
 12666 if (active_lines > 0.0f)
770 12666 reduction = calc_reduction_3gpp(a, desired_pe_no_ah, pe_no_ah, active_lines);
771
772 12666 pe = 0.0f;
773
2/2
✓ Branch 0 taken 13541 times.
✓ Branch 1 taken 12666 times.
 26207 for (w = 0; w < wi->num_windows*16; w += 16) {
774
2/2
✓ Branch 0 taken 628509 times.
✓ Branch 1 taken 13541 times.
 642050 for (g = 0; g < num_bands; g++) {
775 628509 AacPsyBand *band = &pch->band[w+g];
776
777
1/2
✓ Branch 0 taken 628509 times.
✗ Branch 1 not taken.
 628509 if (active_lines > 0.0f)
778 628509 band->thr = calc_reduced_thr_3gpp(band, coeffs[g].min_snr, reduction);
779 628509 pe += calc_pe_3gpp(band);
780
1/2
✓ Branch 0 taken 628509 times.
✗ Branch 1 not taken.
 628509 if (band->thr > 0.0f)
781 628509 band->norm_fac = band->active_lines / band->thr;
782 else
783 ✗ band->norm_fac = 0.0f;
784 628509 norm_fac += band->norm_fac;
785 }
786 }
787 12666 delta_pe = desired_pe - pe;
788
2/2
✓ Branch 0 taken 5878 times.
✓ Branch 1 taken 6788 times.
 12666 if (fabs(delta_pe) > 0.05f * desired_pe)
789 5878 break;
790 }
791
792
2/2
✓ Branch 0 taken 5004 times.
✓ Branch 1 taken 4268 times.
 9272 if (pe < 1.15f * desired_pe) {
793 /* 6.6.1.3.6 "Final threshold modification by linearization" */
794
1/2
✓ Branch 0 taken 5004 times.
✗ Branch 1 not taken.
 5004 norm_fac = norm_fac ? 1.0f / norm_fac : 0;
795
2/2
✓ Branch 0 taken 5004 times.
✓ Branch 1 taken 5004 times.
 10008 for (w = 0; w < wi->num_windows*16; w += 16) {
796
2/2
✓ Branch 0 taken 245196 times.
✓ Branch 1 taken 5004 times.
 250200 for (g = 0; g < num_bands; g++) {
797 245196 AacPsyBand *band = &pch->band[w+g];
798
799
2/2
✓ Branch 0 taken 238349 times.
✓ Branch 1 taken 6847 times.
 245196 if (band->active_lines > 0.5f) {
800 238349 float delta_sfb_pe = band->norm_fac * norm_fac * delta_pe;
801 238349 float thr = band->thr;
802
803 238349 thr *= exp2f(delta_sfb_pe / band->active_lines);
804
4/4
✓ Branch 0 taken 22321 times.
✓ Branch 1 taken 216028 times.
✓ Branch 2 taken 61 times.
✓ Branch 3 taken 22260 times.
 238349 if (thr > coeffs[g].min_snr * band->energy && band->avoid_holes == PSY_3GPP_AH_INACTIVE)
805
1/2
✗ Branch 0 not taken.
✓ Branch 1 taken 61 times.
 61 thr = FFMAX(band->thr, coeffs[g].min_snr * band->energy);
806 238349 band->thr = thr;
807 }
808 }
809 }
810 } else {
811 /* 5.6.1.3.7 "Further perceptual entropy reduction" */
812 4268 g = num_bands;
813
4/4
✓ Branch 0 taken 208641 times.
✓ Branch 1 taken 14 times.
✓ Branch 2 taken 204387 times.
✓ Branch 3 taken 4254 times.
 208655 while (pe > desired_pe && g--) {
814
2/2
✓ Branch 0 taken 216637 times.
✓ Branch 1 taken 204387 times.
 421024 for (w = 0; w < wi->num_windows*16; w+= 16) {
815 216637 AacPsyBand *band = &pch->band[w+g];
816
4/4
✓ Branch 0 taken 216501 times.
✓ Branch 1 taken 136 times.
✓ Branch 2 taken 367 times.
✓ Branch 3 taken 216134 times.
 216637 if (band->avoid_holes != PSY_3GPP_AH_NONE && coeffs[g].min_snr < PSY_SNR_1DB) {
817 367 coeffs[g].min_snr = PSY_SNR_1DB;
818 367 band->thr = band->energy * PSY_SNR_1DB;
819 367 pe += band->active_lines * 1.5f - band->pe;
820 }
821 }
822 }
823 /* TODO: allow more holes (unused without mid/side) */
824 }
825 }
826
827
2/2
✓ Branch 0 taken 10754 times.
✓ Branch 1 taken 9375 times.
 20129 for (w = 0; w < wi->num_windows*16; w += 16) {
828
2/2
✓ Branch 0 taken 471786 times.
✓ Branch 1 taken 10754 times.
 482540 for (g = 0; g < num_bands; g++) {
829 471786 AacPsyBand *band = &pch->band[w+g];
830 471786 FFPsyBand *psy_band = &ctx->ch[channel].psy_bands[w+g];
831
832 471786 psy_band->threshold = band->thr;
833 471786 psy_band->energy = band->energy;
834 471786 psy_band->spread = band->active_lines * 2.0f / band_sizes[g];
835 471786 psy_band->bits = PSY_3GPP_PE_TO_BITS(band->pe);
836 }
837 }
838
839 9375 memcpy(pch->prev_band, pch->band, sizeof(pch->band));
840 9375 }
841
842 5015 static void psy_3gpp_analyze(FFPsyContext *ctx, int channel,
843 const float **coeffs, const FFPsyWindowInfo *wi)
844 {
845 int ch;
846 5015 FFPsyChannelGroup *group = ff_psy_find_group(ctx, channel);
847
848
2/2
✓ Branch 0 taken 9375 times.
✓ Branch 1 taken 5015 times.
 14390 for (ch = 0; ch < group->num_ch; ch++)
849 9375 psy_3gpp_analyze_channel(ctx, channel + ch, coeffs[ch], &wi[ch]);
850 5015 }
851
852 10 static av_cold void psy_3gpp_end(FFPsyContext *apc)
853 {
854 10 AacPsyContext *pctx = (AacPsyContext*) apc->model_priv_data;
855
1/2
✓ Branch 0 taken 10 times.
✗ Branch 1 not taken.
 10 if (pctx)
856 10 av_freep(&pctx->ch);
857 10 av_freep(&apc->model_priv_data);
858 10 }
859
860 6414 static void lame_apply_block_type(AacPsyChannel *ctx, FFPsyWindowInfo *wi, int uselongblock)
861 {
862 6414 int blocktype = ONLY_LONG_SEQUENCE;
863
2/2
✓ Branch 0 taken 6294 times.
✓ Branch 1 taken 120 times.
 6414 if (uselongblock) {
864
2/2
✓ Branch 0 taken 79 times.
✓ Branch 1 taken 6215 times.
 6294 if (ctx->next_window_seq == EIGHT_SHORT_SEQUENCE)
865 79 blocktype = LONG_STOP_SEQUENCE;
866 } else {
867 120 blocktype = EIGHT_SHORT_SEQUENCE;
868
2/2
✓ Branch 0 taken 79 times.
✓ Branch 1 taken 41 times.
 120 if (ctx->next_window_seq == ONLY_LONG_SEQUENCE)
869 79 ctx->next_window_seq = LONG_START_SEQUENCE;
870
2/2
✓ Branch 0 taken 14 times.
✓ Branch 1 taken 106 times.
 120 if (ctx->next_window_seq == LONG_STOP_SEQUENCE)
871 14 ctx->next_window_seq = EIGHT_SHORT_SEQUENCE;
872 }
873
874 6414 wi->window_type[0] = ctx->next_window_seq;
875 6414 ctx->next_window_seq = blocktype;
876 6414 }
877
878 6414 static FFPsyWindowInfo psy_lame_window(FFPsyContext *ctx, const float *audio,
879 const float *la, int channel, int prev_type)
880 {
881 6414 AacPsyContext *pctx = (AacPsyContext*) ctx->model_priv_data;
882 6414 AacPsyChannel *pch = &pctx->ch[channel];
883 6414 int grouping = 0;
884 6414 int uselongblock = 1;
885 6414 int attacks[AAC_NUM_BLOCKS_SHORT + 1] = { 0 };
886 int i;
887 6414 FFPsyWindowInfo wi = { { 0 } };
888
889
2/2
✓ Branch 0 taken 6370 times.
✓ Branch 1 taken 44 times.
 6414 if (la) {
890 float hpfsmpl[AAC_BLOCK_SIZE_LONG];
891 6370 const float *pf = hpfsmpl;
892 float attack_intensity[(AAC_NUM_BLOCKS_SHORT + 1) * PSY_LAME_NUM_SUBBLOCKS];
893 float energy_subshort[(AAC_NUM_BLOCKS_SHORT + 1) * PSY_LAME_NUM_SUBBLOCKS];
894 6370 float energy_short[AAC_NUM_BLOCKS_SHORT + 1] = { 0 };
895 6370 const float *firbuf = la + (AAC_BLOCK_SIZE_SHORT/4 - PSY_LAME_FIR_LEN);
896 6370 int att_sum = 0;
897
898 /* LAME comment: apply high pass filter of fs/4 */
899 6370 psy_hp_filter(firbuf, hpfsmpl, psy_fir_coeffs);
900
901 /* Calculate the energies of each sub-shortblock */
902
2/2
✓ Branch 0 taken 12740 times.
✓ Branch 1 taken 6370 times.
 19110 for (i = 0; i < PSY_LAME_NUM_SUBBLOCKS; i++) {
903 12740 energy_subshort[i] = pch->prev_energy_subshort[i + ((AAC_NUM_BLOCKS_SHORT - 1) * PSY_LAME_NUM_SUBBLOCKS)];
904 assert(pch->prev_energy_subshort[i + ((AAC_NUM_BLOCKS_SHORT - 1) * PSY_LAME_NUM_SUBBLOCKS - 2)] > 0);
905 12740 attack_intensity[i] = energy_subshort[i] / pch->prev_energy_subshort[i + ((AAC_NUM_BLOCKS_SHORT - 1) * PSY_LAME_NUM_SUBBLOCKS - 2)];
906 12740 energy_short[0] += energy_subshort[i];
907 }
908
909
2/2
✓ Branch 0 taken 101920 times.
✓ Branch 1 taken 6370 times.
 108290 for (i = 0; i < AAC_NUM_BLOCKS_SHORT * PSY_LAME_NUM_SUBBLOCKS; i++) {
910 101920 const float *const pfe = pf + AAC_BLOCK_SIZE_LONG / (AAC_NUM_BLOCKS_SHORT * PSY_LAME_NUM_SUBBLOCKS);
911 101920 float p = 1.0f;
912
2/2
✓ Branch 0 taken 6522880 times.
✓ Branch 1 taken 101920 times.
 6624800 for (; pf < pfe; pf++)
913
2/2
✓ Branch 0 taken 5963378 times.
✓ Branch 1 taken 559502 times.
 6522880 p = FFMAX(p, fabsf(*pf));
914 101920 pch->prev_energy_subshort[i] = energy_subshort[i + PSY_LAME_NUM_SUBBLOCKS] = p;
915 101920 energy_short[1 + i / PSY_LAME_NUM_SUBBLOCKS] += p;
916
917 /* NOTE: The indexes below are [i + 3 - 2] in the LAME source. Compare each sub-block to sub-block - 2 */
918
2/2
✓ Branch 0 taken 50944 times.
✓ Branch 1 taken 50976 times.
 101920 if (p > energy_subshort[i + PSY_LAME_NUM_SUBBLOCKS - 2])
919 50944 p = p / energy_subshort[i + PSY_LAME_NUM_SUBBLOCKS - 2];
920
2/2
✓ Branch 0 taken 40 times.
✓ Branch 1 taken 50936 times.
 50976 else if (energy_subshort[i + PSY_LAME_NUM_SUBBLOCKS - 2] > p * 10.0f)
921 40 p = energy_subshort[i + PSY_LAME_NUM_SUBBLOCKS - 2] / (p * 10.0f);
922 else
923 50936 p = 0.0;
924
925 101920 attack_intensity[i + PSY_LAME_NUM_SUBBLOCKS] = p;
926 }
927
928 /* compare energy between sub-short blocks */
929
2/2
✓ Branch 0 taken 114660 times.
✓ Branch 1 taken 6370 times.
 121030 for (i = 0; i < (AAC_NUM_BLOCKS_SHORT + 1) * PSY_LAME_NUM_SUBBLOCKS; i++)
930
2/2
✓ Branch 0 taken 114590 times.
✓ Branch 1 taken 70 times.
 114660 if (!attacks[i / PSY_LAME_NUM_SUBBLOCKS])
931
2/2
✓ Branch 0 taken 118 times.
✓ Branch 1 taken 114472 times.
 114590 if (attack_intensity[i] > pch->attack_threshold)
932 118 attacks[i / PSY_LAME_NUM_SUBBLOCKS] = (i % PSY_LAME_NUM_SUBBLOCKS) + 1;
933
934 /* should have energy change between short blocks, in order to avoid periodic signals */
935 /* Good samples to show the effect are Trumpet test songs */
936 /* GB: tuned (1) to avoid too many short blocks for test sample TRUMPET */
937 /* RH: tuned (2) to let enough short blocks through for test sample FSOL and SNAPS */
938
2/2
✓ Branch 0 taken 50960 times.
✓ Branch 1 taken 6370 times.
 57330 for (i = 1; i < AAC_NUM_BLOCKS_SHORT + 1; i++) {
939 50960 const float u = energy_short[i - 1];
940 50960 const float v = energy_short[i];
941
2/2
✓ Branch 0 taken 25770 times.
✓ Branch 1 taken 25190 times.
 50960 const float m = FFMAX(u, v);
942
2/2
✓ Branch 0 taken 48570 times.
✓ Branch 1 taken 2390 times.
 50960 if (m < 40000) { /* (2) */
943
4/4
✓ Branch 0 taken 48262 times.
✓ Branch 1 taken 308 times.
✓ Branch 2 taken 47765 times.
✓ Branch 3 taken 497 times.
 48570 if (u < 2.3f * v && v < 2.3f * u) { /* (1) */
944
3/4
✓ Branch 0 taken 5918 times.
✓ Branch 1 taken 41847 times.
✗ Branch 2 not taken.
✓ Branch 3 taken 5918 times.
 47765 if (i == 1 && attacks[0] < attacks[i])
945 ✗ attacks[0] = 0;
946 47765 attacks[i] = 0;
947 }
948 }
949 50960 att_sum += attacks[i];
950 }
951
952
2/2
✓ Branch 0 taken 6321 times.
✓ Branch 1 taken 49 times.
 6370 if (pch->next_attack0_zero)
953 6321 attacks[0] = 0;
954 6370 pch->next_attack0_zero = !attacks[AAC_NUM_BLOCKS_SHORT];
955
956
2/2
✓ Branch 0 taken 6344 times.
✓ Branch 1 taken 26 times.
 6370 if (attacks[0] <= pch->prev_attack)
957 6344 attacks[0] = 0;
958
959 6370 att_sum += attacks[0];
960
961 /* If the previous attack happened in the last sub-block of the previous sequence,
962 * or if there's a new attack, use short window */
963
4/4
✓ Branch 0 taken 6369 times.
✓ Branch 1 taken 1 times.
✓ Branch 2 taken 105 times.
✓ Branch 3 taken 6264 times.
 6370 if (pch->prev_attack == PSY_LAME_NUM_SUBBLOCKS || att_sum) {
964 106 uselongblock = 0;
965
966
2/2
✓ Branch 0 taken 848 times.
✓ Branch 1 taken 106 times.
 954 for (i = 1; i < AAC_NUM_BLOCKS_SHORT + 1; i++)
967
4/4
✓ Branch 0 taken 89 times.
✓ Branch 1 taken 759 times.
✓ Branch 2 taken 1 times.
✓ Branch 3 taken 88 times.
 848 if (attacks[i] && attacks[i-1])
968 1 attacks[i] = 0;
969 }
970 } else {
971 /* We have no lookahead info, so just use same type as the previous sequence. */
972 44 uselongblock = !(prev_type == EIGHT_SHORT_SEQUENCE);
973 }
974
975 6414 lame_apply_block_type(pch, &wi, uselongblock);
976
977 6414 wi.window_type[1] = prev_type;
978
2/2
✓ Branch 0 taken 6294 times.
✓ Branch 1 taken 120 times.
 6414 if (wi.window_type[0] != EIGHT_SHORT_SEQUENCE) {
979
980 6294 wi.num_windows = 1;
981 6294 wi.grouping[0] = 1;
982
2/2
✓ Branch 0 taken 79 times.
✓ Branch 1 taken 6215 times.
 6294 if (wi.window_type[0] == LONG_START_SEQUENCE)
983 79 wi.window_shape = 0;
984 else
985 6215 wi.window_shape = 1;
986
987 } else {
988 120 int lastgrp = 0;
989
990 120 wi.num_windows = 8;
991 120 wi.window_shape = 0;
992
2/2
✓ Branch 0 taken 960 times.
✓ Branch 1 taken 120 times.
 1080 for (i = 0; i < 8; i++) {
993
2/2
✓ Branch 0 taken 439 times.
✓ Branch 1 taken 521 times.
 960 if (!((pch->next_grouping >> i) & 1))
994 439 lastgrp = i;
995 960 wi.grouping[lastgrp]++;
996 }
997 }
998
999 /* Determine grouping, based on the location of the first attack, and save for
1000 * the next frame.
1001 * FIXME: Move this to analysis.
1002 * TODO: Tune groupings depending on attack location
1003 * TODO: Handle more than one attack in a group
1004 */
1005
2/2
✓ Branch 0 taken 57224 times.
✓ Branch 1 taken 6309 times.
 63533 for (i = 0; i < 9; i++) {
1006
2/2
✓ Branch 0 taken 105 times.
✓ Branch 1 taken 57119 times.
 57224 if (attacks[i]) {
1007 105 grouping = i;
1008 105 break;
1009 }
1010 }
1011 6414 pch->next_grouping = window_grouping[grouping];
1012
1013 6414 pch->prev_attack = attacks[AAC_NUM_BLOCKS_SHORT - 1];
1014
1015 6414 return wi;
1016 }
1017
1018 const FFPsyModel ff_aac_psy_model =
1019 {
1020 .name = "3GPP TS 26.403-inspired model",
1021 .init = psy_3gpp_init,
1022 .window = psy_lame_window,
1023 .analyze = psy_3gpp_analyze,
1024 .end = psy_3gpp_end,
1025 };
1026