2 * Copyright (c) 2001, 2002 Fabrice Bellard
4 * This file is part of FFmpeg.
6 * FFmpeg is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU Lesser General Public
8 * License as published by the Free Software Foundation; either
9 * version 2.1 of the License, or (at your option) any later version.
11 * FFmpeg is distributed in the hope that it will be useful,
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 * Lesser General Public License for more details.
16 * You should have received a copy of the GNU Lesser General Public
17 * License along with FFmpeg; if not, write to the Free Software
18 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
23 #include "libavutil/mem.h"
26 #include "mpegaudiodsp.h"
27 #include "mpegaudio.h"
30 #define RENAME(n) n##_float
32 static inline float round_sample(float *sum)
39 #define MACS(rt, ra, rb) rt+=(ra)*(rb)
40 #define MULS(ra, rb) ((ra)*(rb))
41 #define MULH3(x, y, s) ((s)*(y)*(x))
42 #define MLSS(rt, ra, rb) rt-=(ra)*(rb)
43 #define MULLx(x, y, s) ((y)*(x))
44 #define FIXHR(x) ((float)(x))
45 #define FIXR(x) ((float)(x))
46 #define SHR(a,b) ((a)*(1.0f/(1<<(b))))
50 #define RENAME(n) n##_fixed
51 #define OUT_SHIFT (WFRAC_BITS + FRAC_BITS - 15)
53 static inline int round_sample(int64_t *sum)
56 sum1 = (int)((*sum) >> OUT_SHIFT);
57 *sum &= (1<<OUT_SHIFT)-1;
58 return av_clip_int16(sum1);
61 # define MULS(ra, rb) MUL64(ra, rb)
62 # define MACS(rt, ra, rb) MAC64(rt, ra, rb)
63 # define MLSS(rt, ra, rb) MLS64(rt, ra, rb)
64 # define MULH3(x, y, s) MULH((s)*(x), y)
65 # define MULLx(x, y, s) MULL(x,y,s)
66 # define SHR(a,b) ((a)>>(b))
67 # define FIXR(a) ((int)((a) * FRAC_ONE + 0.5))
68 # define FIXHR(a) ((int)((a) * (1LL<<32) + 0.5))
71 /** Window for MDCT. Actually only the elements in [0,17] and
72 [MDCT_BUF_SIZE/2, MDCT_BUF_SIZE/2 + 17] are actually used. The rest
73 is just to preserve alignment for SIMD implementations.
75 DECLARE_ALIGNED(16, INTFLOAT, RENAME(ff_mdct_win))[8][MDCT_BUF_SIZE];
77 DECLARE_ALIGNED(16, MPA_INT, RENAME(ff_mpa_synth_window))[512+256];
79 #define SUM8(op, sum, w, p) \
81 op(sum, (w)[0 * 64], (p)[0 * 64]); \
82 op(sum, (w)[1 * 64], (p)[1 * 64]); \
83 op(sum, (w)[2 * 64], (p)[2 * 64]); \
84 op(sum, (w)[3 * 64], (p)[3 * 64]); \
85 op(sum, (w)[4 * 64], (p)[4 * 64]); \
86 op(sum, (w)[5 * 64], (p)[5 * 64]); \
87 op(sum, (w)[6 * 64], (p)[6 * 64]); \
88 op(sum, (w)[7 * 64], (p)[7 * 64]); \
91 #define SUM8P2(sum1, op1, sum2, op2, w1, w2, p) \
95 op1(sum1, (w1)[0 * 64], tmp);\
96 op2(sum2, (w2)[0 * 64], tmp);\
98 op1(sum1, (w1)[1 * 64], tmp);\
99 op2(sum2, (w2)[1 * 64], tmp);\
101 op1(sum1, (w1)[2 * 64], tmp);\
102 op2(sum2, (w2)[2 * 64], tmp);\
104 op1(sum1, (w1)[3 * 64], tmp);\
105 op2(sum2, (w2)[3 * 64], tmp);\
107 op1(sum1, (w1)[4 * 64], tmp);\
108 op2(sum2, (w2)[4 * 64], tmp);\
110 op1(sum1, (w1)[5 * 64], tmp);\
111 op2(sum2, (w2)[5 * 64], tmp);\
113 op1(sum1, (w1)[6 * 64], tmp);\
114 op2(sum2, (w2)[6 * 64], tmp);\
116 op1(sum1, (w1)[7 * 64], tmp);\
117 op2(sum2, (w2)[7 * 64], tmp);\
120 void RENAME(ff_mpadsp_apply_window)(MPA_INT *synth_buf, MPA_INT *window,
121 int *dither_state, OUT_INT *samples,
124 register const MPA_INT *w, *w2, *p;
133 /* copy to avoid wrap */
134 memcpy(synth_buf + 512, synth_buf, 32 * sizeof(*synth_buf));
136 samples2 = samples + 31 * incr;
142 SUM8(MACS, sum, w, p);
144 SUM8(MLSS, sum, w + 32, p);
145 *samples = round_sample(&sum);
149 /* we calculate two samples at the same time to avoid one memory
150 access per two sample */
153 p = synth_buf + 16 + j;
154 SUM8P2(sum, MACS, sum2, MLSS, w, w2, p);
155 p = synth_buf + 48 - j;
156 SUM8P2(sum, MLSS, sum2, MLSS, w + 32, w2 + 32, p);
158 *samples = round_sample(&sum);
161 *samples2 = round_sample(&sum);
168 SUM8(MLSS, sum, w + 32, p);
169 *samples = round_sample(&sum);
173 /* 32 sub band synthesis filter. Input: 32 sub band samples, Output:
175 void RENAME(ff_mpa_synth_filter)(MPADSPContext *s, MPA_INT *synth_buf_ptr,
176 int *synth_buf_offset,
177 MPA_INT *window, int *dither_state,
178 OUT_INT *samples, int incr,
184 offset = *synth_buf_offset;
185 synth_buf = synth_buf_ptr + offset;
187 s->RENAME(dct32)(synth_buf, sb_samples);
188 s->RENAME(apply_window)(synth_buf, window, dither_state, samples, incr);
190 offset = (offset - 32) & 511;
191 *synth_buf_offset = offset;
194 av_cold void RENAME(ff_mpa_synth_init)(MPA_INT *window)
198 /* max = 18760, max sum over all 16 coefs : 44736 */
201 v = ff_mpa_enwindow[i];
203 v *= 1.0 / (1LL<<(16 + FRAC_BITS));
213 // Needed for avoiding shuffles in ASM implementations
215 for(j=0; j < 16; j++)
216 window[512+16*i+j] = window[64*i+32-j];
219 for(j=0; j < 16; j++)
220 window[512+128+16*i+j] = window[64*i+48-j];
223 void RENAME(ff_init_mpadsp_tabs)(void)
226 /* compute mdct windows */
227 for (i = 0; i < 36; i++) {
228 for (j = 0; j < 4; j++) {
231 if (j == 2 && i % 3 != 1)
234 d = sin(M_PI * (i + 0.5) / 36.0);
237 else if (i >= 24) d = sin(M_PI * (i - 18 + 0.5) / 12.0);
238 else if (i >= 18) d = 1;
241 else if (i < 12) d = sin(M_PI * (i - 6 + 0.5) / 12.0);
242 else if (i < 18) d = 1;
244 //merge last stage of imdct into the window coefficients
245 d *= 0.5 * IMDCT_SCALAR / cos(M_PI * (2 * i + 19) / 72);
248 RENAME(ff_mdct_win)[j][i/3] = FIXHR((d / (1<<5)));
250 int idx = i < 18 ? i : i + (MDCT_BUF_SIZE/2 - 18);
251 RENAME(ff_mdct_win)[j][idx] = FIXHR((d / (1<<5)));
256 /* NOTE: we do frequency inversion adter the MDCT by changing
257 the sign of the right window coefs */
258 for (j = 0; j < 4; j++) {
259 for (i = 0; i < MDCT_BUF_SIZE; i += 2) {
260 RENAME(ff_mdct_win)[j + 4][i ] = RENAME(ff_mdct_win)[j][i ];
261 RENAME(ff_mdct_win)[j + 4][i + 1] = -RENAME(ff_mdct_win)[j][i + 1];
266 #define C1 FIXHR(0.98480775301220805936/2)
267 #define C2 FIXHR(0.93969262078590838405/2)
268 #define C3 FIXHR(0.86602540378443864676/2)
269 #define C4 FIXHR(0.76604444311897803520/2)
270 #define C5 FIXHR(0.64278760968653932632/2)
271 #define C6 FIXHR(0.5/2)
272 #define C7 FIXHR(0.34202014332566873304/2)
273 #define C8 FIXHR(0.17364817766693034885/2)
275 /* 0.5 / cos(pi*(2*i+1)/36) */
276 static const INTFLOAT icos36[9] = {
277 FIXR(0.50190991877167369479),
278 FIXR(0.51763809020504152469), //0
279 FIXR(0.55168895948124587824),
280 FIXR(0.61038729438072803416),
281 FIXR(0.70710678118654752439), //1
282 FIXR(0.87172339781054900991),
283 FIXR(1.18310079157624925896),
284 FIXR(1.93185165257813657349), //2
285 FIXR(5.73685662283492756461),
288 /* 0.5 / cos(pi*(2*i+1)/36) */
289 static const INTFLOAT icos36h[9] = {
290 FIXHR(0.50190991877167369479/2),
291 FIXHR(0.51763809020504152469/2), //0
292 FIXHR(0.55168895948124587824/2),
293 FIXHR(0.61038729438072803416/2),
294 FIXHR(0.70710678118654752439/2), //1
295 FIXHR(0.87172339781054900991/2),
296 FIXHR(1.18310079157624925896/4),
297 FIXHR(1.93185165257813657349/4), //2
298 // FIXHR(5.73685662283492756461),
301 /* using Lee like decomposition followed by hand coded 9 points DCT */
302 static void imdct36(INTFLOAT *out, INTFLOAT *buf, INTFLOAT *in, INTFLOAT *win)
305 INTFLOAT t0, t1, t2, t3, s0, s1, s2, s3;
306 INTFLOAT tmp[18], *tmp1, *in1;
308 for (i = 17; i >= 1; i--)
310 for (i = 17; i >= 3; i -= 2)
313 for (j = 0; j < 2; j++) {
317 t2 = in1[2*4] + in1[2*8] - in1[2*2];
319 t3 = in1[2*0] + SHR(in1[2*6],1);
320 t1 = in1[2*0] - in1[2*6];
321 tmp1[ 6] = t1 - SHR(t2,1);
324 t0 = MULH3(in1[2*2] + in1[2*4] , C2, 2);
325 t1 = MULH3(in1[2*4] - in1[2*8] , -2*C8, 1);
326 t2 = MULH3(in1[2*2] + in1[2*8] , -C4, 2);
328 tmp1[10] = t3 - t0 - t2;
329 tmp1[ 2] = t3 + t0 + t1;
330 tmp1[14] = t3 + t2 - t1;
332 tmp1[ 4] = MULH3(in1[2*5] + in1[2*7] - in1[2*1], -C3, 2);
333 t2 = MULH3(in1[2*1] + in1[2*5], C1, 2);
334 t3 = MULH3(in1[2*5] - in1[2*7], -2*C7, 1);
335 t0 = MULH3(in1[2*3], C3, 2);
337 t1 = MULH3(in1[2*1] + in1[2*7], -C5, 2);
339 tmp1[ 0] = t2 + t3 + t0;
340 tmp1[12] = t2 + t1 - t0;
341 tmp1[ 8] = t3 - t1 - t0;
345 for (j = 0; j < 4; j++) {
353 s1 = MULH3(t3 + t2, icos36h[ j], 2);
354 s3 = MULLx(t3 - t2, icos36 [8 - j], FRAC_BITS);
358 out[(9 + j) * SBLIMIT] = MULH3(t1, win[ 9 + j], 1) + buf[4*(9 + j)];
359 out[(8 - j) * SBLIMIT] = MULH3(t1, win[ 8 - j], 1) + buf[4*(8 - j)];
360 buf[4 * ( 9 + j )] = MULH3(t0, win[MDCT_BUF_SIZE/2 + 9 + j], 1);
361 buf[4 * ( 8 - j )] = MULH3(t0, win[MDCT_BUF_SIZE/2 + 8 - j], 1);
365 out[(9 + 8 - j) * SBLIMIT] = MULH3(t1, win[ 9 + 8 - j], 1) + buf[4*(9 + 8 - j)];
366 out[ j * SBLIMIT] = MULH3(t1, win[ j], 1) + buf[4*( j)];
367 buf[4 * ( 9 + 8 - j )] = MULH3(t0, win[MDCT_BUF_SIZE/2 + 9 + 8 - j], 1);
368 buf[4 * ( j )] = MULH3(t0, win[MDCT_BUF_SIZE/2 + j], 1);
373 s1 = MULH3(tmp[17], icos36h[4], 2);
376 out[(9 + 4) * SBLIMIT] = MULH3(t1, win[ 9 + 4], 1) + buf[4*(9 + 4)];
377 out[(8 - 4) * SBLIMIT] = MULH3(t1, win[ 8 - 4], 1) + buf[4*(8 - 4)];
378 buf[4 * ( 9 + 4 )] = MULH3(t0, win[MDCT_BUF_SIZE/2 + 9 + 4], 1);
379 buf[4 * ( 8 - 4 )] = MULH3(t0, win[MDCT_BUF_SIZE/2 + 8 - 4], 1);
382 void RENAME(ff_imdct36_blocks)(INTFLOAT *out, INTFLOAT *buf, INTFLOAT *in,
383 int count, int switch_point, int block_type)
386 for (j=0 ; j < count; j++) {
387 /* apply window & overlap with previous buffer */
390 int win_idx = (switch_point && j < 2) ? 0 : block_type;
391 INTFLOAT *win = RENAME(ff_mdct_win)[win_idx + (4 & -(j & 1))];
393 imdct36(out, buf, in, win);
396 buf += ((j&3) != 3 ? 1 : (72-3));