FFmpeg coverage

Directory: ../../../ffmpeg/
File: src/libavcodec/aacpsy.c
Date: 2025-03-08 20:38:41
Exec Total Coverage
Lines: 369 452 81.6%
Functions: 16 18 88.9%
Branches: 263 347 75.8%
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 3 ///< 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 }AacPsyChannel;
137
138 /**
139 * psychoacoustic model frame type-dependent coefficients
140 */
141 typedef struct AacPsyCoeffs{
142 float ath; ///< absolute threshold of hearing per bands
143 float barks; ///< Bark value for each spectral band in long frame
144 float spread_low[2]; ///< spreading factor for low-to-high threshold spreading in long frame
145 float spread_hi [2]; ///< spreading factor for high-to-low threshold spreading in long frame
146 float min_snr; ///< minimal SNR
147 }AacPsyCoeffs;
148
149 /**
150 * 3GPP TS26.403-inspired psychoacoustic model specific data
151 */
152 typedef struct AacPsyContext{
153 int chan_bitrate; ///< bitrate per channel
154 int frame_bits; ///< average bits per frame
155 int fill_level; ///< bit reservoir fill level
156 struct {
157 float min; ///< minimum allowed PE for bit factor calculation
158 float max; ///< maximum allowed PE for bit factor calculation
159 float previous; ///< allowed PE of the previous frame
160 float correction; ///< PE correction factor
161 } pe;
162 AacPsyCoeffs psy_coef[2][64];
163 AacPsyChannel *ch;
164 float global_quality; ///< normalized global quality taken from avctx
165 }AacPsyContext;
166
167 /**
168 * LAME psy model preset struct
169 */
170 typedef struct PsyLamePreset {
171 int quality; ///< Quality to map the rest of the vaules to.
172 /* This is overloaded to be both kbps per channel in ABR mode, and
173 * requested quality in constant quality mode.
174 */
175 float st_lrm; ///< short threshold for L, R, and M channels
176 } PsyLamePreset;
177
178 /**
179 * LAME psy model preset table for ABR
180 */
181 static const PsyLamePreset psy_abr_map[] = {
182 /* TODO: Tuning. These were taken from LAME. */
183 /* kbps/ch st_lrm */
184 { 8, 6.60},
185 { 16, 6.60},
186 { 24, 6.60},
187 { 32, 6.60},
188 { 40, 6.60},
189 { 48, 6.60},
190 { 56, 6.60},
191 { 64, 6.40},
192 { 80, 6.00},
193 { 96, 5.60},
194 {112, 5.20},
195 {128, 5.20},
196 {160, 5.20}
197 };
198
199 /**
200 * LAME psy model preset table for constant quality
201 */
202 static const PsyLamePreset psy_vbr_map[] = {
203 /* vbr_q st_lrm */
204 { 0, 4.20},
205 { 1, 4.20},
206 { 2, 4.20},
207 { 3, 4.20},
208 { 4, 4.20},
209 { 5, 4.20},
210 { 6, 4.20},
211 { 7, 4.20},
212 { 8, 4.20},
213 { 9, 4.20},
214 {10, 4.20}
215 };
216
217 /**
218 * LAME psy model FIR coefficient table
219 */
220 static const float psy_fir_coeffs[] = {
221 -8.65163e-18 * 2, -0.00851586 * 2, -6.74764e-18 * 2, 0.0209036 * 2,
222 -3.36639e-17 * 2, -0.0438162 * 2, -1.54175e-17 * 2, 0.0931738 * 2,
223 -5.52212e-17 * 2, -0.313819 * 2
224 };
225
226 /**
227 * Calculate the ABR attack threshold from the above LAME psymodel table.
228 */
229 23 static float lame_calc_attack_threshold(int bitrate)
230 {
231 /* Assume max bitrate to start with */
232 23 int lower_range = 12, upper_range = 12;
233 23 int lower_range_kbps = psy_abr_map[12].quality;
234 23 int upper_range_kbps = psy_abr_map[12].quality;
235 int i;
236
237 /* Determine which bitrates the value specified falls between.
238 * If the loop ends without breaking our above assumption of 320kbps was correct.
239 */
240
2/2
✓ Branch 0 taken 194 times.
✓ Branch 1 taken 4 times.
 198 for (i = 1; i < 13; i++) {
241
2/2
✓ Branch 0 taken 19 times.
✓ Branch 1 taken 175 times.
 194 if (FFMAX(bitrate, psy_abr_map[i].quality) != bitrate) {
242 19 upper_range = i;
243 19 upper_range_kbps = psy_abr_map[i ].quality;
244 19 lower_range = i - 1;
245 19 lower_range_kbps = psy_abr_map[i - 1].quality;
246 19 break; /* Upper range found */
247 }
248 }
249
250 /* Determine which range the value specified is closer to */
251
2/2
✓ Branch 0 taken 19 times.
✓ Branch 1 taken 4 times.
 23 if ((upper_range_kbps - bitrate) > (bitrate - lower_range_kbps))
252 19 return psy_abr_map[lower_range].st_lrm;
253 4 return psy_abr_map[upper_range].st_lrm;
254 }
255
256 /**
257 * LAME psy model specific initialization
258 */
259 10 static av_cold void lame_window_init(AacPsyContext *ctx, AVCodecContext *avctx)
260 {
261 int i, j;
262
263
2/2
✓ Branch 0 taken 23 times.
✓ Branch 1 taken 10 times.
 33 for (i = 0; i < avctx->ch_layout.nb_channels; i++) {
264 23 AacPsyChannel *pch = &ctx->ch[i];
265
266
1/2
✗ Branch 0 not taken.
✓ Branch 1 taken 23 times.
 23 if (avctx->flags & AV_CODEC_FLAG_QSCALE)
267 ✗ pch->attack_threshold = psy_vbr_map[av_clip(avctx->global_quality / FF_QP2LAMBDA, 0, 10)].st_lrm;
268 else
269 23 pch->attack_threshold = lame_calc_attack_threshold(avctx->bit_rate / avctx->ch_layout.nb_channels / 1000);
270
271
2/2
✓ Branch 0 taken 552 times.
✓ Branch 1 taken 23 times.
 575 for (j = 0; j < AAC_NUM_BLOCKS_SHORT * PSY_LAME_NUM_SUBBLOCKS; j++)
272 552 pch->prev_energy_subshort[j] = 10.0f;
273 }
274 10 }
275
276 /**
277 * Calculate Bark value for given line.
278 */
279 640 static av_cold float calc_bark(float f)
280 {
281 640 return 13.3f * atanf(0.00076f * f) + 3.5f * atanf((f / 7500.0f) * (f / 7500.0f));
282 }
283
284 #define ATH_ADD 4
285 /**
286 * Calculate ATH value for given frequency.
287 * Borrowed from Lame.
288 */
289 13036 static av_cold float ath(float f, float add)
290 {
291 13036 f /= 1000.0f;
292 13036 return 3.64 * pow(f, -0.8)
293 13036 - 6.8 * exp(-0.6 * (f - 3.4) * (f - 3.4))
294 13036 + 6.0 * exp(-0.15 * (f - 8.7) * (f - 8.7))
295 13036 + (0.6 + 0.04 * add) * 0.001 * f * f * f * f;
296 }
297
298 10 static av_cold int psy_3gpp_init(FFPsyContext *ctx) {
299 AacPsyContext *pctx;
300 float bark;
301 int i, j, g, start;
302 float prev, minscale, minath, minsnr, pe_min;
303
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);
304
305
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);
306 10 const float num_bark = calc_bark((float)bandwidth);
307
308
1/2
✗ Branch 0 not taken.
✓ Branch 1 taken 10 times.
 10 if (bandwidth <= 0)
309 ✗ return AVERROR(EINVAL);
310
311 10 ctx->model_priv_data = av_mallocz(sizeof(AacPsyContext));
312
1/2
✗ Branch 0 not taken.
✓ Branch 1 taken 10 times.
 10 if (!ctx->model_priv_data)
313 ✗ return AVERROR(ENOMEM);
314 10 pctx = ctx->model_priv_data;
315
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;
316
317
1/2
✗ Branch 0 not taken.
✓ Branch 1 taken 10 times.
 10 if (ctx->avctx->flags & AV_CODEC_FLAG_QSCALE) {
318 /* Use the target average bitrate to compute spread parameters */
319 ✗ chan_bitrate = (int)(chan_bitrate / 120.0 * (ctx->avctx->global_quality ? ctx->avctx->global_quality : 120));
320 }
321
322 10 pctx->chan_bitrate = chan_bitrate;
323 10 pctx->frame_bits = FFMIN(2560, chan_bitrate * AAC_BLOCK_SIZE_LONG / ctx->avctx->sample_rate);
324 10 pctx->pe.min = 8.0f * AAC_BLOCK_SIZE_LONG * bandwidth / (ctx->avctx->sample_rate * 2.0f);
325 10 pctx->pe.max = 12.0f * AAC_BLOCK_SIZE_LONG * bandwidth / (ctx->avctx->sample_rate * 2.0f);
326 10 ctx->bitres.size = 6144 - pctx->frame_bits;
327 10 ctx->bitres.size -= ctx->bitres.size % 8;
328 10 pctx->fill_level = ctx->bitres.size;
329 10 minath = ath(3410 - 0.733 * ATH_ADD, ATH_ADD);
330
2/2
✓ Branch 0 taken 20 times.
✓ Branch 1 taken 10 times.
 30 for (j = 0; j < 2; j++) {
331 20 AacPsyCoeffs *coeffs = pctx->psy_coef[j];
332 20 const uint8_t *band_sizes = ctx->bands[j];
333
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);
334
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;
335 /* reference encoder uses 2.4% here instead of 60% like the spec says */
336 20 float bark_pe = 0.024f * PSY_3GPP_BITS_TO_PE(avg_chan_bits) / num_bark;
337
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;
338 /* High energy spreading for long blocks <= 22kbps/channel and short blocks are the same. */
339
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;
340
341 20 i = 0;
342 20 prev = 0.0;
343
2/2
✓ Branch 0 taken 630 times.
✓ Branch 1 taken 20 times.
 650 for (g = 0; g < ctx->num_bands[j]; g++) {
344 630 i += band_sizes[g];
345 630 bark = calc_bark((i-1) * line_to_frequency);
346 630 coeffs[g].barks = (bark + prev) / 2.0;
347 630 prev = bark;
348 }
349
2/2
✓ Branch 0 taken 610 times.
✓ Branch 1 taken 20 times.
 630 for (g = 0; g < ctx->num_bands[j] - 1; g++) {
350 610 AacPsyCoeffs *coeff = &coeffs[g];
351 610 float bark_width = coeffs[g+1].barks - coeffs->barks;
352 610 coeff->spread_low[0] = ff_exp10(-bark_width * PSY_3GPP_THR_SPREAD_LOW);
353 610 coeff->spread_hi [0] = ff_exp10(-bark_width * PSY_3GPP_THR_SPREAD_HI);
354 610 coeff->spread_low[1] = ff_exp10(-bark_width * en_spread_low);
355 610 coeff->spread_hi [1] = ff_exp10(-bark_width * en_spread_hi);
356 610 pe_min = bark_pe * bark_width;
357 610 minsnr = exp2(pe_min / band_sizes[g]) - 1.5f;
358 610 coeff->min_snr = av_clipf(1.0f / minsnr, PSY_SNR_25DB, PSY_SNR_1DB);
359 }
360 20 start = 0;
361
2/2
✓ Branch 0 taken 630 times.
✓ Branch 1 taken 20 times.
 650 for (g = 0; g < ctx->num_bands[j]; g++) {
362 630 minscale = ath(start * line_to_frequency, ATH_ADD);
363
2/2
✓ Branch 0 taken 10890 times.
✓ Branch 1 taken 630 times.
 11520 for (i = 1; i < band_sizes[g]; i++)
364
2/2
✓ Branch 1 taken 1506 times.
✓ Branch 2 taken 9384 times.
 10890 minscale = FFMIN(minscale, ath((start + i) * line_to_frequency, ATH_ADD));
365 630 coeffs[g].ath = minscale - minath;
366 630 start += band_sizes[g];
367 }
368 }
369
370 10 pctx->ch = av_calloc(ctx->avctx->ch_layout.nb_channels, sizeof(*pctx->ch));
371
1/2
✗ Branch 0 not taken.
✓ Branch 1 taken 10 times.
 10 if (!pctx->ch) {
372 ✗ av_freep(&ctx->model_priv_data);
373 ✗ return AVERROR(ENOMEM);
374 }
375
376 10 lame_window_init(pctx, ctx->avctx);
377
378 10 return 0;
379 }
380
381 /**
382 * IIR filter used in block switching decision
383 */
384 ✗ static float iir_filter(int in, float state[2])
385 {
386 float ret;
387
388 ✗ ret = 0.7548f * (in - state[0]) + 0.5095f * state[1];
389 ✗ state[0] = in;
390 ✗ state[1] = ret;
391 ✗ return ret;
392 }
393
394 /**
395 * window grouping information stored as bits (0 - new group, 1 - group continues)
396 */
397 static const uint8_t window_grouping[9] = {
398 0xB6, 0x6C, 0xD8, 0xB2, 0x66, 0xC6, 0x96, 0x36, 0x36
399 };
400
401 /**
402 * Tell encoder which window types to use.
403 * @see 3GPP TS26.403 5.4.1 "Blockswitching"
404 */
405 ✗ static av_unused FFPsyWindowInfo psy_3gpp_window(FFPsyContext *ctx,
406 const int16_t *audio,
407 const int16_t *la,
408 int channel, int prev_type)
409 {
410 int i, j;
411 ✗ int br = ((AacPsyContext*)ctx->model_priv_data)->chan_bitrate;
412 ✗ int attack_ratio = br <= 16000 ? 18 : 10;
413 ✗ AacPsyContext *pctx = (AacPsyContext*) ctx->model_priv_data;
414 ✗ AacPsyChannel *pch = &pctx->ch[channel];
415 ✗ uint8_t grouping = 0;
416 ✗ int next_type = pch->next_window_seq;
417 ✗ FFPsyWindowInfo wi = { { 0 } };
418
419 ✗ if (la) {
420 float s[8], v;
421 ✗ int switch_to_eight = 0;
422 ✗ float sum = 0.0, sum2 = 0.0;
423 ✗ int attack_n = 0;
424 ✗ int stay_short = 0;
425 ✗ for (i = 0; i < 8; i++) {
426 ✗ for (j = 0; j < 128; j++) {
427 ✗ v = iir_filter(la[i*128+j], pch->iir_state);
428 ✗ sum += v*v;
429 }
430 ✗ s[i] = sum;
431 ✗ sum2 += sum;
432 }
433 ✗ for (i = 0; i < 8; i++) {
434 ✗ if (s[i] > pch->win_energy * attack_ratio) {
435 ✗ attack_n = i + 1;
436 ✗ switch_to_eight = 1;
437 ✗ break;
438 }
439 }
440 ✗ pch->win_energy = pch->win_energy*7/8 + sum2/64;
441
442 ✗ wi.window_type[1] = prev_type;
443 ✗ switch (prev_type) {
444 ✗ case ONLY_LONG_SEQUENCE:
445 ✗ wi.window_type[0] = switch_to_eight ? LONG_START_SEQUENCE : ONLY_LONG_SEQUENCE;
446 ✗ next_type = switch_to_eight ? EIGHT_SHORT_SEQUENCE : ONLY_LONG_SEQUENCE;
447 ✗ break;
448 ✗ case LONG_START_SEQUENCE:
449 ✗ wi.window_type[0] = EIGHT_SHORT_SEQUENCE;
450 ✗ grouping = pch->next_grouping;
451 ✗ next_type = switch_to_eight ? EIGHT_SHORT_SEQUENCE : LONG_STOP_SEQUENCE;
452 ✗ break;
453 ✗ case LONG_STOP_SEQUENCE:
454 ✗ wi.window_type[0] = switch_to_eight ? LONG_START_SEQUENCE : ONLY_LONG_SEQUENCE;
455 ✗ next_type = switch_to_eight ? EIGHT_SHORT_SEQUENCE : ONLY_LONG_SEQUENCE;
456 ✗ break;
457 ✗ case EIGHT_SHORT_SEQUENCE:
458 ✗ stay_short = next_type == EIGHT_SHORT_SEQUENCE || switch_to_eight;
459 ✗ wi.window_type[0] = stay_short ? EIGHT_SHORT_SEQUENCE : LONG_STOP_SEQUENCE;
460 ✗ grouping = next_type == EIGHT_SHORT_SEQUENCE ? pch->next_grouping : 0;
461 ✗ next_type = switch_to_eight ? EIGHT_SHORT_SEQUENCE : LONG_STOP_SEQUENCE;
462 ✗ break;
463 }
464
465 ✗ pch->next_grouping = window_grouping[attack_n];
466 ✗ pch->next_window_seq = next_type;
467 } else {
468 ✗ for (i = 0; i < 3; i++)
469 ✗ wi.window_type[i] = prev_type;
470 ✗ grouping = (prev_type == EIGHT_SHORT_SEQUENCE) ? window_grouping[0] : 0;
471 }
472
473 ✗ wi.window_shape = 1;
474 ✗ if (wi.window_type[0] != EIGHT_SHORT_SEQUENCE) {
475 ✗ wi.num_windows = 1;
476 ✗ wi.grouping[0] = 1;
477 } else {
478 ✗ int lastgrp = 0;
479 ✗ wi.num_windows = 8;
480 ✗ for (i = 0; i < 8; i++) {
481 ✗ if (!((grouping >> i) & 1))
482 ✗ lastgrp = i;
483 ✗ wi.grouping[lastgrp]++;
484 }
485 }
486
487 ✗ return wi;
488 }
489
490 /* 5.6.1.2 "Calculation of Bit Demand" */
491 9399 static int calc_bit_demand(AacPsyContext *ctx, float pe, int bits, int size,
492 int short_window)
493 {
494
2/2
✓ Branch 0 taken 224 times.
✓ Branch 1 taken 9175 times.
 9399 const float bitsave_slope = short_window ? PSY_3GPP_SAVE_SLOPE_S : PSY_3GPP_SAVE_SLOPE_L;
495
2/2
✓ Branch 0 taken 224 times.
✓ Branch 1 taken 9175 times.
 9399 const float bitsave_add = short_window ? PSY_3GPP_SAVE_ADD_S : PSY_3GPP_SAVE_ADD_L;
496
2/2
✓ Branch 0 taken 224 times.
✓ Branch 1 taken 9175 times.
 9399 const float bitspend_slope = short_window ? PSY_3GPP_SPEND_SLOPE_S : PSY_3GPP_SPEND_SLOPE_L;
497
2/2
✓ Branch 0 taken 224 times.
✓ Branch 1 taken 9175 times.
 9399 const float bitspend_add = short_window ? PSY_3GPP_SPEND_ADD_S : PSY_3GPP_SPEND_ADD_L;
498 9399 const float clip_low = short_window ? PSY_3GPP_CLIP_LO_S : PSY_3GPP_CLIP_LO_L;
499
2/2
✓ Branch 0 taken 224 times.
✓ Branch 1 taken 9175 times.
 9399 const float clip_high = short_window ? PSY_3GPP_CLIP_HI_S : PSY_3GPP_CLIP_HI_L;
500 float clipped_pe, bit_save, bit_spend, bit_factor, fill_level, forgetful_min_pe;
501
502 9399 ctx->fill_level += ctx->frame_bits - bits;
503 9399 ctx->fill_level = av_clip(ctx->fill_level, 0, size);
504 9399 fill_level = av_clipf((float)ctx->fill_level / size, clip_low, clip_high);
505 9399 clipped_pe = av_clipf(pe, ctx->pe.min, ctx->pe.max);
506 9399 bit_save = (fill_level + bitsave_add) * bitsave_slope;
507 assert(bit_save <= 0.3f && bit_save >= -0.05000001f);
508 9399 bit_spend = (fill_level + bitspend_add) * bitspend_slope;
509 assert(bit_spend <= 0.5f && bit_spend >= -0.1f);
510 /* The bit factor graph in the spec is obviously incorrect.
511 * bit_spend + ((bit_spend - bit_spend))...
512 * The reference encoder subtracts everything from 1, but also seems incorrect.
513 * 1 - bit_save + ((bit_spend + bit_save))...
514 * Hopefully below is correct.
515 */
516 9399 bit_factor = 1.0f - bit_save + ((bit_spend - bit_save) / (ctx->pe.max - ctx->pe.min)) * (clipped_pe - ctx->pe.min);
517 /* NOTE: The reference encoder attempts to center pe max/min around the current pe.
518 * Here we do that by slowly forgetting pe.min when pe stays in a range that makes
519 * it unlikely (ie: above the mean)
520 */
521
2/2
✓ Branch 0 taken 19 times.
✓ Branch 1 taken 9380 times.
 9399 ctx->pe.max = FFMAX(pe, ctx->pe.max);
522 18798 forgetful_min_pe = ((ctx->pe.min * PSY_PE_FORGET_SLOPE)
523
2/2
✓ Branch 0 taken 2775 times.
✓ Branch 1 taken 6624 times.
 9399 + FFMAX(ctx->pe.min, pe * (pe / ctx->pe.max))) / (PSY_PE_FORGET_SLOPE + 1);
524
2/2
✓ Branch 0 taken 9315 times.
✓ Branch 1 taken 84 times.
 9399 ctx->pe.min = FFMIN(pe, forgetful_min_pe);
525
526 /* NOTE: allocate a minimum of 1/8th average frame bits, to avoid
527 * reservoir starvation from producing zero-bit frames
528 */
529
6/6
✓ Branch 0 taken 9385 times.
✓ Branch 1 taken 14 times.
✓ Branch 2 taken 1160 times.
✓ Branch 3 taken 8239 times.
✓ Branch 4 taken 1146 times.
✓ Branch 5 taken 14 times.
 9399 return FFMIN(
530 ctx->frame_bits * bit_factor,
531 FFMAX(ctx->frame_bits + size - bits, ctx->frame_bits / 8));
532 }
533
534 1568756 static float calc_pe_3gpp(AacPsyBand *band)
535 {
536 float pe, a;
537
538 1568756 band->pe = 0.0f;
539 1568756 band->pe_const = 0.0f;
540 1568756 band->active_lines = 0.0f;
541
2/2
✓ Branch 0 taken 1522192 times.
✓ Branch 1 taken 46564 times.
 1568756 if (band->energy > band->thr) {
542 1522192 a = log2f(band->energy);
543 1522192 pe = a - log2f(band->thr);
544 1522192 band->active_lines = band->nz_lines;
545
2/2
✓ Branch 0 taken 729838 times.
✓ Branch 1 taken 792354 times.
 1522192 if (pe < PSY_3GPP_C1) {
546 729838 pe = pe * PSY_3GPP_C3 + PSY_3GPP_C2;
547 729838 a = a * PSY_3GPP_C3 + PSY_3GPP_C2;
548 729838 band->active_lines *= PSY_3GPP_C3;
549 }
550 1522192 band->pe = pe * band->nz_lines;
551 1522192 band->pe_const = a * band->nz_lines;
552 }
553
554 1568756 return band->pe;
555 }
556
557 22975 static float calc_reduction_3gpp(float a, float desired_pe, float pe,
558 float active_lines)
559 {
560 float thr_avg, reduction;
561
562
2/2
✓ Branch 0 taken 24 times.
✓ Branch 1 taken 22951 times.
 22975 if(active_lines == 0.0)
563 24 return 0;
564
565 22951 thr_avg = exp2f((a - pe) / (4.0f * active_lines));
566 22951 reduction = exp2f((a - desired_pe) / (4.0f * active_lines)) - thr_avg;
567
568
2/2
✓ Branch 0 taken 20891 times.
✓ Branch 1 taken 2060 times.
 22951 return FFMAX(reduction, 0.0f);
569 }
570
571 1094093 static float calc_reduced_thr_3gpp(AacPsyBand *band, float min_snr,
572 float reduction)
573 {
574 1094093 float thr = band->thr;
575
576
2/2
✓ Branch 0 taken 1067422 times.
✓ Branch 1 taken 26671 times.
 1094093 if (band->energy > thr) {
577 1067422 thr = sqrtf(thr);
578 1067422 thr = sqrtf(thr) + reduction;
579 1067422 thr *= thr;
580 1067422 thr *= thr;
581
582 /* This deviates from the 3GPP spec to match the reference encoder.
583 * It performs min(thr_reduced, max(thr, energy/min_snr)) only for bands
584 * that have hole avoidance on (active or inactive). It always reduces the
585 * threshold of bands with hole avoidance off.
586 */
587
4/4
✓ Branch 0 taken 335795 times.
✓ Branch 1 taken 731627 times.
✓ Branch 2 taken 335727 times.
✓ Branch 3 taken 68 times.
 1067422 if (thr > band->energy * min_snr && band->avoid_holes != PSY_3GPP_AH_NONE) {
588
2/2
✓ Branch 0 taken 4684 times.
✓ Branch 1 taken 331043 times.
 335727 thr = FFMAX(band->thr, band->energy * min_snr);
589 335727 band->avoid_holes = PSY_3GPP_AH_ACTIVE;
590 }
591 }
592
593 1094093 return thr;
594 }
595
596 #ifndef calc_thr_3gpp
597 9399 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 9399 int start = 0, wstart = 0;
602
2/2
✓ Branch 0 taken 10967 times.
✓ Branch 1 taken 9399 times.
 20366 for (w = 0; w < wi->num_windows*16; w += 16) {
603 10967 wstart = 0;
604
2/2
✓ Branch 0 taken 474663 times.
✓ Branch 1 taken 10967 times.
 485630 for (g = 0; g < num_bands; g++) {
605 474663 AacPsyBand *band = &pch->band[w+g];
606
607 474663 float form_factor = 0.0f;
608 float Temp;
609 474663 band->energy = 0.0f;
610
2/2
✓ Branch 0 taken 466912 times.
✓ Branch 1 taken 7751 times.
 474663 if (wstart < cutoff) {
611
2/2
✓ Branch 0 taken 9307424 times.
✓ Branch 1 taken 466912 times.
 9774336 for (i = 0; i < band_sizes[g]; i++) {
612 9307424 band->energy += coefs[start+i] * coefs[start+i];
613 9307424 form_factor += sqrtf(fabs(coefs[start+i]));
614 }
615 }
616
2/2
✓ Branch 0 taken 457427 times.
✓ Branch 1 taken 17236 times.
 474663 Temp = band->energy > 0 ? sqrtf((float)band_sizes[g] / band->energy) : 0;
617 474663 band->thr = band->energy * 0.001258925f;
618 474663 band->nz_lines = form_factor * sqrtf(Temp);
619
620 474663 start += band_sizes[g];
621 474663 wstart += band_sizes[g];
622 }
623 }
624 9399 }
625 #endif /* calc_thr_3gpp */
626
627 #ifndef psy_hp_filter
628 6370 static void psy_hp_filter(const float *firbuf, float *hpfsmpl, const float *psy_fir_coeffs)
629 {
630 int i, j;
631
2/2
✓ Branch 0 taken 6522880 times.
✓ Branch 1 taken 6370 times.
 6529250 for (i = 0; i < AAC_BLOCK_SIZE_LONG; i++) {
632 float sum1, sum2;
633 6522880 sum1 = firbuf[i + (PSY_LAME_FIR_LEN - 1) / 2];
634 6522880 sum2 = 0.0;
635
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) {
636 32614400 sum1 += psy_fir_coeffs[j] * (firbuf[i + j] + firbuf[i + PSY_LAME_FIR_LEN - j]);
637 32614400 sum2 += psy_fir_coeffs[j + 1] * (firbuf[i + j + 1] + firbuf[i + PSY_LAME_FIR_LEN - j - 1]);
638 }
639 /* NOTE: The LAME psymodel expects it's input in the range -32768 to 32768.
640 * Tuning this for normalized floats would be difficult. */
641 6522880 hpfsmpl[i] = (sum1 + sum2) * 32768.0f;
642 }
643 6370 }
644 #endif /* psy_hp_filter */
645
646 /**
647 * Calculate band thresholds as suggested in 3GPP TS26.403
648 */
649 9399 static void psy_3gpp_analyze_channel(FFPsyContext *ctx, int channel,
650 const float *coefs, const FFPsyWindowInfo *wi)
651 {
652 9399 AacPsyContext *pctx = (AacPsyContext*) ctx->model_priv_data;
653 9399 AacPsyChannel *pch = &pctx->ch[channel];
654 int i, w, g;
655 9399 float desired_bits, desired_pe, delta_pe, reduction= NAN, spread_en[128] = {0};
656 9399 float a = 0.0f, active_lines = 0.0f, norm_fac = 0.0f;
657
1/4
✗ Branch 0 not taken.
✓ Branch 1 taken 9399 times.
✗ Branch 2 not taken.
✗ Branch 3 not taken.
 9399 float pe = pctx->chan_bitrate > 32000 ? 0.0f : FFMAX(50.0f, 100.0f - pctx->chan_bitrate * 100.0f / 32000.0f);
658
2/2
✓ Branch 0 taken 224 times.
✓ Branch 1 taken 9175 times.
 9399 const int num_bands = ctx->num_bands[wi->num_windows == 8];
659
2/2
✓ Branch 0 taken 224 times.
✓ Branch 1 taken 9175 times.
 9399 const uint8_t *band_sizes = ctx->bands[wi->num_windows == 8];
660 9399 AacPsyCoeffs *coeffs = pctx->psy_coef[wi->num_windows == 8];
661
2/2
✓ Branch 0 taken 224 times.
✓ Branch 1 taken 9175 times.
 9399 const float avoid_hole_thr = wi->num_windows == 8 ? PSY_3GPP_AH_THR_SHORT : PSY_3GPP_AH_THR_LONG;
662
5/8
✓ Branch 0 taken 4904 times.
✓ Branch 1 taken 4495 times.
✗ Branch 2 not taken.
✓ Branch 3 taken 4495 times.
✓ Branch 4 taken 4495 times.
✗ Branch 5 not taken.
✗ Branch 6 not taken.
✓ Branch 7 taken 4495 times.
 9399 const int bandwidth = ctx->cutoff ? ctx->cutoff : AAC_CUTOFF(ctx->avctx);
663 9399 const int cutoff = bandwidth * 2048 / wi->num_windows / ctx->avctx->sample_rate;
664
665 //calculate energies, initial thresholds and related values - 5.4.2 "Threshold Calculation"
666 9399 calc_thr_3gpp(wi, num_bands, pch, band_sizes, coefs, cutoff);
667
668 //modify thresholds and energies - spread, threshold in quiet, pre-echo control
669
2/2
✓ Branch 0 taken 10967 times.
✓ Branch 1 taken 9399 times.
 20366 for (w = 0; w < wi->num_windows*16; w += 16) {
670 10967 AacPsyBand *bands = &pch->band[w];
671
672 /* 5.4.2.3 "Spreading" & 5.4.3 "Spread Energy Calculation" */
673 10967 spread_en[0] = bands[0].energy;
674
2/2
✓ Branch 0 taken 463696 times.
✓ Branch 1 taken 10967 times.
 474663 for (g = 1; g < num_bands; g++) {
675
2/2
✓ Branch 0 taken 446532 times.
✓ Branch 1 taken 17164 times.
 463696 bands[g].thr = FFMAX(bands[g].thr, bands[g-1].thr * coeffs[g].spread_hi[0]);
676
2/2
✓ Branch 0 taken 447019 times.
✓ Branch 1 taken 16677 times.
 463696 spread_en[w+g] = FFMAX(bands[g].energy, spread_en[w+g-1] * coeffs[g].spread_hi[1]);
677 }
678
2/2
✓ Branch 0 taken 463696 times.
✓ Branch 1 taken 10967 times.
 474663 for (g = num_bands - 2; g >= 0; g--) {
679
2/2
✓ Branch 0 taken 449647 times.
✓ Branch 1 taken 14049 times.
 463696 bands[g].thr = FFMAX(bands[g].thr, bands[g+1].thr * coeffs[g].spread_low[0]);
680
2/2
✓ Branch 0 taken 447511 times.
✓ Branch 1 taken 16185 times.
 463696 spread_en[w+g] = FFMAX(spread_en[w+g], spread_en[w+g+1] * coeffs[g].spread_low[1]);
681 }
682 //5.4.2.4 "Threshold in quiet"
683
2/2
✓ Branch 0 taken 474663 times.
✓ Branch 1 taken 10967 times.
 485630 for (g = 0; g < num_bands; g++) {
684 474663 AacPsyBand *band = &bands[g];
685
686
2/2
✓ Branch 0 taken 346692 times.
✓ Branch 1 taken 127971 times.
 474663 band->thr_quiet = band->thr = FFMAX(band->thr, coeffs[g].ath);
687 //5.4.2.5 "Pre-echo control"
688
6/6
✓ Branch 0 taken 469518 times.
✓ Branch 1 taken 5145 times.
✓ Branch 2 taken 447566 times.
✓ Branch 3 taken 21952 times.
✓ Branch 4 taken 445900 times.
✓ Branch 5 taken 1666 times.
 474663 if (!(wi->window_type[0] == LONG_STOP_SEQUENCE || (!w && wi->window_type[1] == LONG_START_SEQUENCE)))
689
6/6
✓ Branch 0 taken 53907 times.
✓ Branch 1 taken 413945 times.
✓ Branch 2 taken 4468 times.
✓ Branch 3 taken 463384 times.
✓ Branch 4 taken 49439 times.
✓ Branch 5 taken 413945 times.
 467852 band->thr = FFMAX(PSY_3GPP_RPEMIN*band->thr, FFMIN(band->thr,
690 PSY_3GPP_RPELEV*pch->prev_band[w+g].thr_quiet));
691
692 /* 5.6.1.3.1 "Preparatory steps of the perceptual entropy calculation" */
693 474663 pe += calc_pe_3gpp(band);
694 474663 a += band->pe_const;
695 474663 active_lines += band->active_lines;
696
697 /* 5.6.1.3.3 "Selection of the bands for avoidance of holes" */
698
3/4
✓ Branch 0 taken 474460 times.
✓ Branch 1 taken 203 times.
✗ Branch 2 not taken.
✓ Branch 3 taken 474460 times.
 474663 if (spread_en[w+g] * avoid_hole_thr > band->energy || coeffs[g].min_snr > 1.0f)
699 203 band->avoid_holes = PSY_3GPP_AH_NONE;
700 else
701 474460 band->avoid_holes = PSY_3GPP_AH_INACTIVE;
702 }
703 }
704
705 /* 5.6.1.3.2 "Calculation of the desired perceptual entropy" */
706 9399 ctx->ch[channel].entropy = pe;
707
1/2
✗ Branch 0 not taken.
✓ Branch 1 taken 9399 times.
 9399 if (ctx->avctx->flags & AV_CODEC_FLAG_QSCALE) {
708 /* (2.5 * 120) achieves almost transparent rate, and we want to give
709 * ample room downwards, so we make that equivalent to QSCALE=2.4
710 */
711 ✗ desired_pe = pe * (ctx->avctx->global_quality ? ctx->avctx->global_quality : 120) / (2 * 2.5f * 120.0f);
712 ✗ desired_bits = FFMIN(2560, PSY_3GPP_PE_TO_BITS(desired_pe));
713 ✗ desired_pe = PSY_3GPP_BITS_TO_PE(desired_bits); // reflect clipping
714
715 /* PE slope smoothing */
716 ✗ if (ctx->bitres.bits > 0) {
717 ✗ desired_bits = FFMIN(2560, PSY_3GPP_PE_TO_BITS(desired_pe));
718 ✗ desired_pe = PSY_3GPP_BITS_TO_PE(desired_bits); // reflect clipping
719 }
720
721 ✗ pctx->pe.max = FFMAX(pe, pctx->pe.max);
722 ✗ pctx->pe.min = FFMIN(pe, pctx->pe.min);
723 } else {
724 9399 desired_bits = calc_bit_demand(pctx, pe, ctx->bitres.bits, ctx->bitres.size, wi->num_windows == 8);
725 9399 desired_pe = PSY_3GPP_BITS_TO_PE(desired_bits);
726
727 /* NOTE: PE correction is kept simple. During initial testing it had very
728 * little effect on the final bitrate. Probably a good idea to come
729 * back and do more testing later.
730 */
731
2/2
✓ Branch 0 taken 9338 times.
✓ Branch 1 taken 61 times.
 9399 if (ctx->bitres.bits > 0)
732 9338 desired_pe *= av_clipf(pctx->pe.previous / PSY_3GPP_BITS_TO_PE(ctx->bitres.bits),
733 0.85f, 1.15f);
734 }
735 9399 pctx->pe.previous = PSY_3GPP_BITS_TO_PE(desired_bits);
736 9399 ctx->bitres.alloc = desired_bits;
737
738
2/2
✓ Branch 0 taken 9290 times.
✓ Branch 1 taken 109 times.
 9399 if (desired_pe < pe) {
739 /* 5.6.1.3.4 "First Estimation of the reduction value" */
740
2/2
✓ Branch 0 taken 10312 times.
✓ Branch 1 taken 9290 times.
 19602 for (w = 0; w < wi->num_windows*16; w += 16) {
741 10312 reduction = calc_reduction_3gpp(a, desired_pe, pe, active_lines);
742 10312 pe = 0.0f;
743 10312 a = 0.0f;
744 10312 active_lines = 0.0f;
745
2/2
✓ Branch 0 taken 464408 times.
✓ Branch 1 taken 10312 times.
 474720 for (g = 0; g < num_bands; g++) {
746 464408 AacPsyBand *band = &pch->band[w+g];
747
748 464408 band->thr = calc_reduced_thr_3gpp(band, coeffs[g].min_snr, reduction);
749 /* recalculate PE */
750 464408 pe += calc_pe_3gpp(band);
751 464408 a += band->pe_const;
752 464408 active_lines += band->active_lines;
753 }
754 }
755
756 /* 5.6.1.3.5 "Second Estimation of the reduction value" */
757
2/2
✓ Branch 0 taken 12663 times.
✓ Branch 1 taken 3373 times.
 16036 for (i = 0; i < 2; i++) {
758 12663 float pe_no_ah = 0.0f, desired_pe_no_ah;
759 12663 active_lines = a = 0.0f;
760
2/2
✓ Branch 0 taken 13685 times.
✓ Branch 1 taken 12663 times.
 26348 for (w = 0; w < wi->num_windows*16; w += 16) {
761
2/2
✓ Branch 0 taken 629685 times.
✓ Branch 1 taken 13685 times.
 643370 for (g = 0; g < num_bands; g++) {
762 629685 AacPsyBand *band = &pch->band[w+g];
763
764
2/2
✓ Branch 0 taken 502734 times.
✓ Branch 1 taken 126951 times.
 629685 if (band->avoid_holes != PSY_3GPP_AH_ACTIVE) {
765 502734 pe_no_ah += band->pe;
766 502734 a += band->pe_const;
767 502734 active_lines += band->active_lines;
768 }
769 }
770 }
771
2/2
✓ Branch 0 taken 12627 times.
✓ Branch 1 taken 36 times.
 12663 desired_pe_no_ah = FFMAX(desired_pe - (pe - pe_no_ah), 0.0f);
772
1/2
✓ Branch 0 taken 12663 times.
✗ Branch 1 not taken.
 12663 if (active_lines > 0.0f)
773 12663 reduction = calc_reduction_3gpp(a, desired_pe_no_ah, pe_no_ah, active_lines);
774
775 12663 pe = 0.0f;
776
2/2
✓ Branch 0 taken 13685 times.
✓ Branch 1 taken 12663 times.
 26348 for (w = 0; w < wi->num_windows*16; w += 16) {
777
2/2
✓ Branch 0 taken 629685 times.
✓ Branch 1 taken 13685 times.
 643370 for (g = 0; g < num_bands; g++) {
778 629685 AacPsyBand *band = &pch->band[w+g];
779
780
1/2
✓ Branch 0 taken 629685 times.
✗ Branch 1 not taken.
 629685 if (active_lines > 0.0f)
781 629685 band->thr = calc_reduced_thr_3gpp(band, coeffs[g].min_snr, reduction);
782 629685 pe += calc_pe_3gpp(band);
783
1/2
✓ Branch 0 taken 629685 times.
✗ Branch 1 not taken.
 629685 if (band->thr > 0.0f)
784 629685 band->norm_fac = band->active_lines / band->thr;
785 else
786 ✗ band->norm_fac = 0.0f;
787 629685 norm_fac += band->norm_fac;
788 }
789 }
790 12663 delta_pe = desired_pe - pe;
791
2/2
✓ Branch 0 taken 5917 times.
✓ Branch 1 taken 6746 times.
 12663 if (fabs(delta_pe) > 0.05f * desired_pe)
792 5917 break;
793 }
794
795
2/2
✓ Branch 0 taken 5065 times.
✓ Branch 1 taken 4225 times.
 9290 if (pe < 1.15f * desired_pe) {
796 /* 6.6.1.3.6 "Final threshold modification by linearization" */
797
1/2
✓ Branch 0 taken 5065 times.
✗ Branch 1 not taken.
 5065 norm_fac = norm_fac ? 1.0f / norm_fac : 0;
798
2/2
✓ Branch 0 taken 5065 times.
✓ Branch 1 taken 5065 times.
 10130 for (w = 0; w < wi->num_windows*16; w += 16) {
799
2/2
✓ Branch 0 taken 248185 times.
✓ Branch 1 taken 5065 times.
 253250 for (g = 0; g < num_bands; g++) {
800 248185 AacPsyBand *band = &pch->band[w+g];
801
802
2/2
✓ Branch 0 taken 241414 times.
✓ Branch 1 taken 6771 times.
 248185 if (band->active_lines > 0.5f) {
803 241414 float delta_sfb_pe = band->norm_fac * norm_fac * delta_pe;
804 241414 float thr = band->thr;
805
806 241414 thr *= exp2f(delta_sfb_pe / band->active_lines);
807
4/4
✓ Branch 0 taken 23009 times.
✓ Branch 1 taken 218405 times.
✓ Branch 2 taken 66 times.
✓ Branch 3 taken 22943 times.
 241414 if (thr > coeffs[g].min_snr * band->energy && band->avoid_holes == PSY_3GPP_AH_INACTIVE)
808
1/2
✗ Branch 0 not taken.
✓ Branch 1 taken 66 times.
 66 thr = FFMAX(band->thr, coeffs[g].min_snr * band->energy);
809 241414 band->thr = thr;
810 }
811 }
812 }
813 } else {
814 /* 5.6.1.3.7 "Further perceptual entropy reduction" */
815 4225 g = num_bands;
816
4/4
✓ Branch 0 taken 205756 times.
✓ Branch 1 taken 14 times.
✓ Branch 2 taken 201545 times.
✓ Branch 3 taken 4211 times.
 205770 while (pe > desired_pe && g--) {
817
2/2
✓ Branch 0 taken 215853 times.
✓ Branch 1 taken 201545 times.
 417398 for (w = 0; w < wi->num_windows*16; w+= 16) {
818 215853 AacPsyBand *band = &pch->band[w+g];
819
4/4
✓ Branch 0 taken 215709 times.
✓ Branch 1 taken 144 times.
✓ Branch 2 taken 367 times.
✓ Branch 3 taken 215342 times.
 215853 if (band->avoid_holes != PSY_3GPP_AH_NONE && coeffs[g].min_snr < PSY_SNR_1DB) {
820 367 coeffs[g].min_snr = PSY_SNR_1DB;
821 367 band->thr = band->energy * PSY_SNR_1DB;
822 367 pe += band->active_lines * 1.5f - band->pe;
823 }
824 }
825 }
826 /* TODO: allow more holes (unused without mid/side) */
827 }
828 }
829
830
2/2
✓ Branch 0 taken 10967 times.
✓ Branch 1 taken 9399 times.
 20366 for (w = 0; w < wi->num_windows*16; w += 16) {
831
2/2
✓ Branch 0 taken 474663 times.
✓ Branch 1 taken 10967 times.
 485630 for (g = 0; g < num_bands; g++) {
832 474663 AacPsyBand *band = &pch->band[w+g];
833 474663 FFPsyBand *psy_band = &ctx->ch[channel].psy_bands[w+g];
834
835 474663 psy_band->threshold = band->thr;
836 474663 psy_band->energy = band->energy;
837 474663 psy_band->spread = band->active_lines * 2.0f / band_sizes[g];
838 474663 psy_band->bits = PSY_3GPP_PE_TO_BITS(band->pe);
839 }
840 }
841
842 9399 memcpy(pch->prev_band, pch->band, sizeof(pch->band));
843 9399 }
844
845 5032 static void psy_3gpp_analyze(FFPsyContext *ctx, int channel,
846 const float **coeffs, const FFPsyWindowInfo *wi)
847 {
848 int ch;
849 5032 FFPsyChannelGroup *group = ff_psy_find_group(ctx, channel);
850
851
2/2
✓ Branch 0 taken 9399 times.
✓ Branch 1 taken 5032 times.
 14431 for (ch = 0; ch < group->num_ch; ch++)
852 9399 psy_3gpp_analyze_channel(ctx, channel + ch, coeffs[ch], &wi[ch]);
853 5032 }
854
855 10 static av_cold void psy_3gpp_end(FFPsyContext *apc)
856 {
857 10 AacPsyContext *pctx = (AacPsyContext*) apc->model_priv_data;
858
1/2
✓ Branch 0 taken 10 times.
✗ Branch 1 not taken.
 10 if (pctx)
859 10 av_freep(&pctx->ch);
860 10 av_freep(&apc->model_priv_data);
861 10 }
862
863 6414 static void lame_apply_block_type(AacPsyChannel *ctx, FFPsyWindowInfo *wi, int uselongblock)
864 {
865 6414 int blocktype = ONLY_LONG_SEQUENCE;
866
2/2
✓ Branch 0 taken 6270 times.
✓ Branch 1 taken 144 times.
 6414 if (uselongblock) {
867
2/2
✓ Branch 0 taken 106 times.
✓ Branch 1 taken 6164 times.
 6270 if (ctx->next_window_seq == EIGHT_SHORT_SEQUENCE)
868 106 blocktype = LONG_STOP_SEQUENCE;
869 } else {
870 144 blocktype = EIGHT_SHORT_SEQUENCE;
871
2/2
✓ Branch 0 taken 105 times.
✓ Branch 1 taken 39 times.
 144 if (ctx->next_window_seq == ONLY_LONG_SEQUENCE)
872 105 ctx->next_window_seq = LONG_START_SEQUENCE;
873
2/2
✓ Branch 0 taken 16 times.
✓ Branch 1 taken 128 times.
 144 if (ctx->next_window_seq == LONG_STOP_SEQUENCE)
874 16 ctx->next_window_seq = EIGHT_SHORT_SEQUENCE;
875 }
876
877 6414 wi->window_type[0] = ctx->next_window_seq;
878 6414 ctx->next_window_seq = blocktype;
879 6414 }
880
881 6414 static FFPsyWindowInfo psy_lame_window(FFPsyContext *ctx, const float *audio,
882 const float *la, int channel, int prev_type)
883 {
884 6414 AacPsyContext *pctx = (AacPsyContext*) ctx->model_priv_data;
885 6414 AacPsyChannel *pch = &pctx->ch[channel];
886 6414 int grouping = 0;
887 6414 int uselongblock = 1;
888 6414 int attacks[AAC_NUM_BLOCKS_SHORT + 1] = { 0 };
889 int i;
890 6414 FFPsyWindowInfo wi = { { 0 } };
891
892
2/2
✓ Branch 0 taken 6370 times.
✓ Branch 1 taken 44 times.
 6414 if (la) {
893 float hpfsmpl[AAC_BLOCK_SIZE_LONG];
894 6370 const float *pf = hpfsmpl;
895 float attack_intensity[(AAC_NUM_BLOCKS_SHORT + 1) * PSY_LAME_NUM_SUBBLOCKS];
896 float energy_subshort[(AAC_NUM_BLOCKS_SHORT + 1) * PSY_LAME_NUM_SUBBLOCKS];
897 6370 float energy_short[AAC_NUM_BLOCKS_SHORT + 1] = { 0 };
898 6370 const float *firbuf = la + (AAC_BLOCK_SIZE_SHORT/4 - PSY_LAME_FIR_LEN);
899 6370 int att_sum = 0;
900
901 /* LAME comment: apply high pass filter of fs/4 */
902 6370 psy_hp_filter(firbuf, hpfsmpl, psy_fir_coeffs);
903
904 /* Calculate the energies of each sub-shortblock */
905
2/2
✓ Branch 0 taken 19110 times.
✓ Branch 1 taken 6370 times.
 25480 for (i = 0; i < PSY_LAME_NUM_SUBBLOCKS; i++) {
906 19110 energy_subshort[i] = pch->prev_energy_subshort[i + ((AAC_NUM_BLOCKS_SHORT - 1) * PSY_LAME_NUM_SUBBLOCKS)];
907 assert(pch->prev_energy_subshort[i + ((AAC_NUM_BLOCKS_SHORT - 2) * PSY_LAME_NUM_SUBBLOCKS + 1)] > 0);
908 19110 attack_intensity[i] = energy_subshort[i] / pch->prev_energy_subshort[i + ((AAC_NUM_BLOCKS_SHORT - 2) * PSY_LAME_NUM_SUBBLOCKS + 1)];
909 19110 energy_short[0] += energy_subshort[i];
910 }
911
912
2/2
✓ Branch 0 taken 152880 times.
✓ Branch 1 taken 6370 times.
 159250 for (i = 0; i < AAC_NUM_BLOCKS_SHORT * PSY_LAME_NUM_SUBBLOCKS; i++) {
913 152880 const float *const pfe = pf + AAC_BLOCK_SIZE_LONG / (AAC_NUM_BLOCKS_SHORT * PSY_LAME_NUM_SUBBLOCKS);
914 152880 float p = 1.0f;
915
2/2
✓ Branch 0 taken 6420960 times.
✓ Branch 1 taken 152880 times.
 6573840 for (; pf < pfe; pf++)
916
2/2
✓ Branch 0 taken 5673974 times.
✓ Branch 1 taken 746986 times.
 6420960 p = FFMAX(p, fabsf(*pf));
917 152880 pch->prev_energy_subshort[i] = energy_subshort[i + PSY_LAME_NUM_SUBBLOCKS] = p;
918 152880 energy_short[1 + i / PSY_LAME_NUM_SUBBLOCKS] += p;
919 /* NOTE: The indexes below are [i + 3 - 2] in the LAME source.
920 * Obviously the 3 and 2 have some significance, or this would be just [i + 1]
921 * (which is what we use here). What the 3 stands for is ambiguous, as it is both
922 * number of short blocks, and the number of sub-short blocks.
923 * It seems that LAME is comparing each sub-block to sub-block + 1 in the
924 * previous block.
925 */
926
2/2
✓ Branch 0 taken 75711 times.
✓ Branch 1 taken 77169 times.
 152880 if (p > energy_subshort[i + 1])
927 75711 p = p / energy_subshort[i + 1];
928
2/2
✓ Branch 0 taken 47 times.
✓ Branch 1 taken 77122 times.
 77169 else if (energy_subshort[i + 1] > p * 10.0f)
929 47 p = energy_subshort[i + 1] / (p * 10.0f);
930 else
931 77122 p = 0.0;
932 152880 attack_intensity[i + PSY_LAME_NUM_SUBBLOCKS] = p;
933 }
934
935 /* compare energy between sub-short blocks */
936
2/2
✓ Branch 0 taken 171990 times.
✓ Branch 1 taken 6370 times.
 178360 for (i = 0; i < (AAC_NUM_BLOCKS_SHORT + 1) * PSY_LAME_NUM_SUBBLOCKS; i++)
937
2/2
✓ Branch 0 taken 171825 times.
✓ Branch 1 taken 165 times.
 171990 if (!attacks[i / PSY_LAME_NUM_SUBBLOCKS])
938
2/2
✓ Branch 0 taken 169 times.
✓ Branch 1 taken 171656 times.
 171825 if (attack_intensity[i] > pch->attack_threshold)
939 169 attacks[i / PSY_LAME_NUM_SUBBLOCKS] = (i % PSY_LAME_NUM_SUBBLOCKS) + 1;
940
941 /* should have energy change between short blocks, in order to avoid periodic signals */
942 /* Good samples to show the effect are Trumpet test songs */
943 /* GB: tuned (1) to avoid too many short blocks for test sample TRUMPET */
944 /* RH: tuned (2) to let enough short blocks through for test sample FSOL and SNAPS */
945
2/2
✓ Branch 0 taken 50960 times.
✓ Branch 1 taken 6370 times.
 57330 for (i = 1; i < AAC_NUM_BLOCKS_SHORT + 1; i++) {
946 50960 const float u = energy_short[i - 1];
947 50960 const float v = energy_short[i];
948
2/2
✓ Branch 0 taken 25901 times.
✓ Branch 1 taken 25059 times.
 50960 const float m = FFMAX(u, v);
949
2/2
✓ Branch 0 taken 47964 times.
✓ Branch 1 taken 2996 times.
 50960 if (m < 40000) { /* (2) */
950
4/4
✓ Branch 0 taken 46705 times.
✓ Branch 1 taken 1259 times.
✓ Branch 2 taken 45159 times.
✓ Branch 3 taken 1546 times.
 47964 if (u < 1.7f * v && v < 1.7f * u) { /* (1) */
951
4/4
✓ Branch 0 taken 5611 times.
✓ Branch 1 taken 39548 times.
✓ Branch 2 taken 8 times.
✓ Branch 3 taken 5603 times.
 45159 if (i == 1 && attacks[0] < attacks[i])
952 8 attacks[0] = 0;
953 45159 attacks[i] = 0;
954 }
955 }
956 50960 att_sum += attacks[i];
957 }
958
959
1/2
✓ Branch 0 taken 6370 times.
✗ Branch 1 not taken.
 6370 if (attacks[0] <= pch->prev_attack)
960 6370 attacks[0] = 0;
961
962 6370 att_sum += attacks[0];
963 /* 3 below indicates the previous attack happened in the last sub-block of the previous sequence */
964
4/4
✓ Branch 0 taken 6358 times.
✓ Branch 1 taken 12 times.
✓ Branch 2 taken 117 times.
✓ Branch 3 taken 6241 times.
 6370 if (pch->prev_attack == 3 || att_sum) {
965 129 uselongblock = 0;
966
967
2/2
✓ Branch 0 taken 1032 times.
✓ Branch 1 taken 129 times.
 1161 for (i = 1; i < AAC_NUM_BLOCKS_SHORT + 1; i++)
968
3/4
✓ Branch 0 taken 127 times.
✓ Branch 1 taken 905 times.
✗ Branch 2 not taken.
✓ Branch 3 taken 127 times.
 1032 if (attacks[i] && attacks[i-1])
969 ✗ attacks[i] = 0;
970 }
971 } else {
972 /* We have no lookahead info, so just use same type as the previous sequence. */
973 44 uselongblock = !(prev_type == EIGHT_SHORT_SEQUENCE);
974 }
975
976 6414 lame_apply_block_type(pch, &wi, uselongblock);
977
978 6414 wi.window_type[1] = prev_type;
979
2/2
✓ Branch 0 taken 6269 times.
✓ Branch 1 taken 145 times.
 6414 if (wi.window_type[0] != EIGHT_SHORT_SEQUENCE) {
980
981 6269 wi.num_windows = 1;
982 6269 wi.grouping[0] = 1;
983
2/2
✓ Branch 0 taken 105 times.
✓ Branch 1 taken 6164 times.
 6269 if (wi.window_type[0] == LONG_START_SEQUENCE)
984 105 wi.window_shape = 0;
985 else
986 6164 wi.window_shape = 1;
987
988 } else {
989 145 int lastgrp = 0;
990
991 145 wi.num_windows = 8;
992 145 wi.window_shape = 0;
993
2/2
✓ Branch 0 taken 1160 times.
✓ Branch 1 taken 145 times.
 1305 for (i = 0; i < 8; i++) {
994
2/2
✓ Branch 0 taken 552 times.
✓ Branch 1 taken 608 times.
 1160 if (!((pch->next_grouping >> i) & 1))
995 552 lastgrp = i;
996 1160 wi.grouping[lastgrp]++;
997 }
998 }
999
1000 /* Determine grouping, based on the location of the first attack, and save for
1001 * the next frame.
1002 * FIXME: Move this to analysis.
1003 * TODO: Tune groupings depending on attack location
1004 * TODO: Handle more than one attack in a group
1005 */
1006
2/2
✓ Branch 0 taken 57367 times.
✓ Branch 1 taken 6297 times.
 63664 for (i = 0; i < 9; i++) {
1007
2/2
✓ Branch 0 taken 117 times.
✓ Branch 1 taken 57250 times.
 57367 if (attacks[i]) {
1008 117 grouping = i;
1009 117 break;
1010 }
1011 }
1012 6414 pch->next_grouping = window_grouping[grouping];
1013
1014 6414 pch->prev_attack = attacks[8];
1015
1016 6414 return wi;
1017 }
1018
1019 const FFPsyModel ff_aac_psy_model =
1020 {
1021 .name = "3GPP TS 26.403-inspired model",
1022 .init = psy_3gpp_init,
1023 .window = psy_lame_window,
1024 .analyze = psy_3gpp_analyze,
1025 .end = psy_3gpp_end,
1026 };
1027