3 * Copyright (c) 2011 Justin Ruggles
5 * This file is part of FFmpeg.
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.
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.
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
22 #include "libavutil/mem_internal.h"
29 static void ac3_exponent_min_c(uint8_t *exp, int num_reuse_blocks, int nb_coefs)
33 if (!num_reuse_blocks)
36 for (i = 0; i < nb_coefs; i++) {
37 uint8_t min_exp = *exp;
38 uint8_t *exp1 = exp + 256;
39 for (blk = 0; blk < num_reuse_blocks; blk++) {
40 uint8_t next_exp = *exp1;
41 if (next_exp < min_exp)
49 static void float_to_fixed24_c(int32_t *dst, const float *src, unsigned int len)
51 const float scale = 1 << 24;
53 *dst++ = lrintf(*src++ * scale);
54 *dst++ = lrintf(*src++ * scale);
55 *dst++ = lrintf(*src++ * scale);
56 *dst++ = lrintf(*src++ * scale);
57 *dst++ = lrintf(*src++ * scale);
58 *dst++ = lrintf(*src++ * scale);
59 *dst++ = lrintf(*src++ * scale);
60 *dst++ = lrintf(*src++ * scale);
65 static void ac3_bit_alloc_calc_bap_c(int16_t *mask, int16_t *psd,
67 int snr_offset, int floor,
68 const uint8_t *bap_tab, uint8_t *bap)
70 int bin, band, band_end;
72 /* special case, if snr offset is -960, set all bap's to zero */
73 if (snr_offset == -960) {
74 memset(bap, 0, AC3_MAX_COEFS);
79 band = ff_ac3_bin_to_band_tab[start];
81 int m = (FFMAX(mask[band] - snr_offset - floor, 0) & 0x1FE0) + floor;
82 band_end = ff_ac3_band_start_tab[++band];
83 band_end = FFMIN(band_end, end);
85 for (; bin < band_end; bin++) {
86 int address = av_clip_uintp2((psd[bin] - m) >> 5, 6);
87 bap[bin] = bap_tab[address];
89 } while (end > band_end);
92 static void ac3_update_bap_counts_c(uint16_t mant_cnt[16], uint8_t *bap,
99 DECLARE_ALIGNED(16, const uint16_t, ff_ac3_bap_bits)[16] = {
100 0, 0, 0, 3, 0, 4, 5, 6, 7, 8, 9, 10, 11, 12, 14, 16
103 static int ac3_compute_mantissa_size_c(uint16_t mant_cnt[6][16])
108 for (blk = 0; blk < AC3_MAX_BLOCKS; blk++) {
109 // bap=1 : 3 mantissas in 5 bits
110 bits += (mant_cnt[blk][1] / 3) * 5;
111 // bap=2 : 3 mantissas in 7 bits
112 // bap=4 : 2 mantissas in 7 bits
113 bits += ((mant_cnt[blk][2] / 3) + (mant_cnt[blk][4] >> 1)) * 7;
114 // bap=3 : 1 mantissa in 3 bits
115 bits += mant_cnt[blk][3] * 3;
116 // bap=5 to 15 : get bits per mantissa from table
117 for (bap = 5; bap < 16; bap++)
118 bits += mant_cnt[blk][bap] * ff_ac3_bap_bits[bap];
123 static void ac3_extract_exponents_c(uint8_t *exp, int32_t *coef, int nb_coefs)
127 for (i = 0; i < nb_coefs; i++) {
128 int v = abs(coef[i]);
129 exp[i] = v ? 23 - av_log2(v) : 24;
133 static void ac3_sum_square_butterfly_int32_c(int64_t sum[4],
134 const int32_t *coef0,
135 const int32_t *coef1,
140 sum[0] = sum[1] = sum[2] = sum[3] = 0;
142 for (i = 0; i < len; i++) {
147 MAC64(sum[0], lt, lt);
148 MAC64(sum[1], rt, rt);
149 MAC64(sum[2], md, md);
150 MAC64(sum[3], sd, sd);
154 static void ac3_sum_square_butterfly_float_c(float sum[4],
161 sum[0] = sum[1] = sum[2] = sum[3] = 0;
163 for (i = 0; i < len; i++) {
175 static void ac3_downmix_5_to_2_symmetric_c(float **samples, float **matrix,
180 float front_mix = matrix[0][0];
181 float center_mix = matrix[0][1];
182 float surround_mix = matrix[0][3];
184 for (i = 0; i < len; i++) {
185 v0 = samples[0][i] * front_mix +
186 samples[1][i] * center_mix +
187 samples[3][i] * surround_mix;
189 v1 = samples[1][i] * center_mix +
190 samples[2][i] * front_mix +
191 samples[4][i] * surround_mix;
198 static void ac3_downmix_5_to_1_symmetric_c(float **samples, float **matrix,
202 float front_mix = matrix[0][0];
203 float center_mix = matrix[0][1];
204 float surround_mix = matrix[0][3];
206 for (i = 0; i < len; i++) {
207 samples[0][i] = samples[0][i] * front_mix +
208 samples[1][i] * center_mix +
209 samples[2][i] * front_mix +
210 samples[3][i] * surround_mix +
211 samples[4][i] * surround_mix;
215 static void ac3_downmix_c(float **samples, float **matrix,
216 int out_ch, int in_ch, int len)
222 for (i = 0; i < len; i++) {
224 for (j = 0; j < in_ch; j++) {
225 v0 += samples[j][i] * matrix[0][j];
226 v1 += samples[j][i] * matrix[1][j];
231 } else if (out_ch == 1) {
232 for (i = 0; i < len; i++) {
234 for (j = 0; j < in_ch; j++)
235 v0 += samples[j][i] * matrix[0][j];
241 static void ac3_downmix_5_to_2_symmetric_c_fixed(int32_t **samples, int16_t **matrix,
246 int16_t front_mix = matrix[0][0];
247 int16_t center_mix = matrix[0][1];
248 int16_t surround_mix = matrix[0][3];
250 for (i = 0; i < len; i++) {
251 v0 = (int64_t)samples[0][i] * front_mix +
252 (int64_t)samples[1][i] * center_mix +
253 (int64_t)samples[3][i] * surround_mix;
255 v1 = (int64_t)samples[1][i] * center_mix +
256 (int64_t)samples[2][i] * front_mix +
257 (int64_t)samples[4][i] * surround_mix;
259 samples[0][i] = (v0+2048)>>12;
260 samples[1][i] = (v1+2048)>>12;
264 static void ac3_downmix_5_to_1_symmetric_c_fixed(int32_t **samples, int16_t **matrix,
269 int16_t front_mix = matrix[0][0];
270 int16_t center_mix = matrix[0][1];
271 int16_t surround_mix = matrix[0][3];
273 for (i = 0; i < len; i++) {
274 v0 = (int64_t)samples[0][i] * front_mix +
275 (int64_t)samples[1][i] * center_mix +
276 (int64_t)samples[2][i] * front_mix +
277 (int64_t)samples[3][i] * surround_mix +
278 (int64_t)samples[4][i] * surround_mix;
280 samples[0][i] = (v0+2048)>>12;
284 static void ac3_downmix_c_fixed(int32_t **samples, int16_t **matrix,
285 int out_ch, int in_ch, int len)
290 for (i = 0; i < len; i++) {
292 for (j = 0; j < in_ch; j++) {
293 v0 += (int64_t)samples[j][i] * matrix[0][j];
294 v1 += (int64_t)samples[j][i] * matrix[1][j];
296 samples[0][i] = (v0+2048)>>12;
297 samples[1][i] = (v1+2048)>>12;
299 } else if (out_ch == 1) {
300 for (i = 0; i < len; i++) {
302 for (j = 0; j < in_ch; j++)
303 v0 += (int64_t)samples[j][i] * matrix[0][j];
304 samples[0][i] = (v0+2048)>>12;
309 void ff_ac3dsp_downmix_fixed(AC3DSPContext *c, int32_t **samples, int16_t **matrix,
310 int out_ch, int in_ch, int len)
312 if (c->in_channels != in_ch || c->out_channels != out_ch) {
313 c->in_channels = in_ch;
314 c->out_channels = out_ch;
315 c->downmix_fixed = NULL;
317 if (in_ch == 5 && out_ch == 2 &&
318 !(matrix[1][0] | matrix[0][2] |
319 matrix[1][3] | matrix[0][4] |
320 (matrix[0][1] ^ matrix[1][1]) |
321 (matrix[0][0] ^ matrix[1][2]))) {
322 c->downmix_fixed = ac3_downmix_5_to_2_symmetric_c_fixed;
323 } else if (in_ch == 5 && out_ch == 1 &&
324 matrix[0][0] == matrix[0][2] &&
325 matrix[0][3] == matrix[0][4]) {
326 c->downmix_fixed = ac3_downmix_5_to_1_symmetric_c_fixed;
330 if (c->downmix_fixed)
331 c->downmix_fixed(samples, matrix, len);
333 ac3_downmix_c_fixed(samples, matrix, out_ch, in_ch, len);
336 void ff_ac3dsp_downmix(AC3DSPContext *c, float **samples, float **matrix,
337 int out_ch, int in_ch, int len)
339 if (c->in_channels != in_ch || c->out_channels != out_ch) {
340 int **matrix_cmp = (int **)matrix;
342 c->in_channels = in_ch;
343 c->out_channels = out_ch;
346 if (in_ch == 5 && out_ch == 2 &&
347 !(matrix_cmp[1][0] | matrix_cmp[0][2] |
348 matrix_cmp[1][3] | matrix_cmp[0][4] |
349 (matrix_cmp[0][1] ^ matrix_cmp[1][1]) |
350 (matrix_cmp[0][0] ^ matrix_cmp[1][2]))) {
351 c->downmix = ac3_downmix_5_to_2_symmetric_c;
352 } else if (in_ch == 5 && out_ch == 1 &&
353 matrix_cmp[0][0] == matrix_cmp[0][2] &&
354 matrix_cmp[0][3] == matrix_cmp[0][4]) {
355 c->downmix = ac3_downmix_5_to_1_symmetric_c;
359 ff_ac3dsp_set_downmix_x86(c);
363 c->downmix(samples, matrix, len);
365 ac3_downmix_c(samples, matrix, out_ch, in_ch, len);
368 av_cold void ff_ac3dsp_init(AC3DSPContext *c, int bit_exact)
370 c->ac3_exponent_min = ac3_exponent_min_c;
371 c->float_to_fixed24 = float_to_fixed24_c;
372 c->bit_alloc_calc_bap = ac3_bit_alloc_calc_bap_c;
373 c->update_bap_counts = ac3_update_bap_counts_c;
374 c->compute_mantissa_size = ac3_compute_mantissa_size_c;
375 c->extract_exponents = ac3_extract_exponents_c;
376 c->sum_square_butterfly_int32 = ac3_sum_square_butterfly_int32_c;
377 c->sum_square_butterfly_float = ac3_sum_square_butterfly_float_c;
381 c->downmix_fixed = NULL;
384 ff_ac3dsp_init_arm(c, bit_exact);
386 ff_ac3dsp_init_x86(c, bit_exact);
388 ff_ac3dsp_init_mips(c, bit_exact);