FFmpeg
 All Data Structures Namespaces Files Functions Variables Typedefs Enumerations Enumerator Macros Groups Pages
fft-test.c
Go to the documentation of this file.
1 /*
2  * (c) 2002 Fabrice Bellard
3  *
4  * This file is part of FFmpeg.
5  *
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.
10  *
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.
15  *
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
19  */
20 
21 /**
22  * @file
23  * FFT and MDCT tests.
24  */
25 
26 #include "libavutil/cpu.h"
27 #include "libavutil/mathematics.h"
28 #include "libavutil/lfg.h"
29 #include "libavutil/log.h"
30 #include "libavutil/time.h"
31 #include "fft.h"
32 #if CONFIG_FFT_FLOAT
33 #include "dct.h"
34 #include "rdft.h"
35 #endif
36 #include <math.h>
37 #if HAVE_UNISTD_H
38 #include <unistd.h>
39 #endif
40 #include <stdio.h>
41 #include <stdlib.h>
42 #include <string.h>
43 
44 /* reference fft */
45 
46 #define MUL16(a,b) ((a) * (b))
47 
48 #define CMAC(pre, pim, are, aim, bre, bim) \
49 {\
50  pre += (MUL16(are, bre) - MUL16(aim, bim));\
51  pim += (MUL16(are, bim) + MUL16(bre, aim));\
52 }
53 
54 #if CONFIG_FFT_FLOAT
55 # define RANGE 1.0
56 # define REF_SCALE(x, bits) (x)
57 # define FMT "%10.6f"
58 #elif CONFIG_FFT_FIXED_32
59 # define RANGE 8388608
60 # define REF_SCALE(x, bits) (x)
61 # define FMT "%6d"
62 #else
63 # define RANGE 16384
64 # define REF_SCALE(x, bits) ((x) / (1<<(bits)))
65 # define FMT "%6d"
66 #endif
67 
68 struct {
69  float re, im;
70 } *exptab;
71 
72 static void fft_ref_init(int nbits, int inverse)
73 {
74  int n, i;
75  double c1, s1, alpha;
76 
77  n = 1 << nbits;
78  exptab = av_malloc((n / 2) * sizeof(*exptab));
79 
80  for (i = 0; i < (n/2); i++) {
81  alpha = 2 * M_PI * (float)i / (float)n;
82  c1 = cos(alpha);
83  s1 = sin(alpha);
84  if (!inverse)
85  s1 = -s1;
86  exptab[i].re = c1;
87  exptab[i].im = s1;
88  }
89 }
90 
91 static void fft_ref(FFTComplex *tabr, FFTComplex *tab, int nbits)
92 {
93  int n, i, j, k, n2;
94  double tmp_re, tmp_im, s, c;
95  FFTComplex *q;
96 
97  n = 1 << nbits;
98  n2 = n >> 1;
99  for (i = 0; i < n; i++) {
100  tmp_re = 0;
101  tmp_im = 0;
102  q = tab;
103  for (j = 0; j < n; j++) {
104  k = (i * j) & (n - 1);
105  if (k >= n2) {
106  c = -exptab[k - n2].re;
107  s = -exptab[k - n2].im;
108  } else {
109  c = exptab[k].re;
110  s = exptab[k].im;
111  }
112  CMAC(tmp_re, tmp_im, c, s, q->re, q->im);
113  q++;
114  }
115  tabr[i].re = REF_SCALE(tmp_re, nbits);
116  tabr[i].im = REF_SCALE(tmp_im, nbits);
117  }
118 }
119 
120 static void imdct_ref(FFTSample *out, FFTSample *in, int nbits)
121 {
122  int n = 1<<nbits;
123  int k, i, a;
124  double sum, f;
125 
126  for (i = 0; i < n; i++) {
127  sum = 0;
128  for (k = 0; k < n/2; k++) {
129  a = (2 * i + 1 + (n / 2)) * (2 * k + 1);
130  f = cos(M_PI * a / (double)(2 * n));
131  sum += f * in[k];
132  }
133  out[i] = REF_SCALE(-sum, nbits - 2);
134  }
135 }
136 
137 /* NOTE: no normalisation by 1 / N is done */
138 static void mdct_ref(FFTSample *output, FFTSample *input, int nbits)
139 {
140  int n = 1<<nbits;
141  int k, i;
142  double a, s;
143 
144  /* do it by hand */
145  for (k = 0; k < n/2; k++) {
146  s = 0;
147  for (i = 0; i < n; i++) {
148  a = (2*M_PI*(2*i+1+n/2)*(2*k+1) / (4 * n));
149  s += input[i] * cos(a);
150  }
151  output[k] = REF_SCALE(s, nbits - 1);
152  }
153 }
154 
155 #if CONFIG_FFT_FLOAT
156 static void idct_ref(FFTSample *output, FFTSample *input, int nbits)
157 {
158  int n = 1<<nbits;
159  int k, i;
160  double a, s;
161 
162  /* do it by hand */
163  for (i = 0; i < n; i++) {
164  s = 0.5 * input[0];
165  for (k = 1; k < n; k++) {
166  a = M_PI*k*(i+0.5) / n;
167  s += input[k] * cos(a);
168  }
169  output[i] = 2 * s / n;
170  }
171 }
172 static void dct_ref(FFTSample *output, FFTSample *input, int nbits)
173 {
174  int n = 1<<nbits;
175  int k, i;
176  double a, s;
177 
178  /* do it by hand */
179  for (k = 0; k < n; k++) {
180  s = 0;
181  for (i = 0; i < n; i++) {
182  a = M_PI*k*(i+0.5) / n;
183  s += input[i] * cos(a);
184  }
185  output[k] = s;
186  }
187 }
188 #endif
189 
190 
191 static FFTSample frandom(AVLFG *prng)
192 {
193  return (int16_t)av_lfg_get(prng) / 32768.0 * RANGE;
194 }
195 
196 static int check_diff(FFTSample *tab1, FFTSample *tab2, int n, double scale)
197 {
198  int i;
199  double max= 0;
200  double error= 0;
201  int err = 0;
202 
203  for (i = 0; i < n; i++) {
204  double e = fabsf(tab1[i] - (tab2[i] / scale)) / RANGE;
205  if (e >= 1e-3) {
206  av_log(NULL, AV_LOG_ERROR, "ERROR %5d: "FMT" "FMT"\n",
207  i, tab1[i], tab2[i]);
208  err = 1;
209  }
210  error+= e*e;
211  if(e>max) max= e;
212  }
213  av_log(NULL, AV_LOG_INFO, "max:%f e:%g\n", max, sqrt(error/n));
214  return err;
215 }
216 
217 
218 static void help(void)
219 {
220  av_log(NULL, AV_LOG_INFO,"usage: fft-test [-h] [-s] [-i] [-n b]\n"
221  "-h print this help\n"
222  "-s speed test\n"
223  "-m (I)MDCT test\n"
224  "-d (I)DCT test\n"
225  "-r (I)RDFT test\n"
226  "-i inverse transform test\n"
227  "-n b set the transform size to 2^b\n"
228  "-f x set scale factor for output data of (I)MDCT to x\n"
229  );
230 }
231 
237 };
238 
239 #if !HAVE_GETOPT
240 #include "compat/getopt.c"
241 #endif
242 
243 int main(int argc, char **argv)
244 {
245  FFTComplex *tab, *tab1, *tab_ref;
246  FFTSample *tab2;
247  int it, i, c;
248  int cpuflags;
249  int do_speed = 0;
250  int err = 1;
252  int do_inverse = 0;
253  FFTContext s1, *s = &s1;
254  FFTContext m1, *m = &m1;
255 #if CONFIG_FFT_FLOAT
256  RDFTContext r1, *r = &r1;
257  DCTContext d1, *d = &d1;
258  int fft_size_2;
259 #endif
260  int fft_nbits, fft_size;
261  double scale = 1.0;
262  AVLFG prng;
263  av_lfg_init(&prng, 1);
264 
265  fft_nbits = 9;
266  for(;;) {
267  c = getopt(argc, argv, "hsimrdn:f:c:");
268  if (c == -1)
269  break;
270  switch(c) {
271  case 'h':
272  help();
273  return 1;
274  case 's':
275  do_speed = 1;
276  break;
277  case 'i':
278  do_inverse = 1;
279  break;
280  case 'm':
281  transform = TRANSFORM_MDCT;
282  break;
283  case 'r':
284  transform = TRANSFORM_RDFT;
285  break;
286  case 'd':
287  transform = TRANSFORM_DCT;
288  break;
289  case 'n':
290  fft_nbits = atoi(optarg);
291  break;
292  case 'f':
293  scale = atof(optarg);
294  break;
295  case 'c':
296  cpuflags = av_get_cpu_flags();
297 
298  if (av_parse_cpu_caps(&cpuflags, optarg) < 0)
299  return 1;
300 
301  av_force_cpu_flags(cpuflags);
302  break;
303  }
304  }
305 
306  fft_size = 1 << fft_nbits;
307  tab = av_malloc(fft_size * sizeof(FFTComplex));
308  tab1 = av_malloc(fft_size * sizeof(FFTComplex));
309  tab_ref = av_malloc(fft_size * sizeof(FFTComplex));
310  tab2 = av_malloc(fft_size * sizeof(FFTSample));
311 
312  switch (transform) {
313  case TRANSFORM_MDCT:
314  av_log(NULL, AV_LOG_INFO,"Scale factor is set to %f\n", scale);
315  if (do_inverse)
316  av_log(NULL, AV_LOG_INFO,"IMDCT");
317  else
318  av_log(NULL, AV_LOG_INFO,"MDCT");
319  ff_mdct_init(m, fft_nbits, do_inverse, scale);
320  break;
321  case TRANSFORM_FFT:
322  if (do_inverse)
323  av_log(NULL, AV_LOG_INFO,"IFFT");
324  else
325  av_log(NULL, AV_LOG_INFO,"FFT");
326  ff_fft_init(s, fft_nbits, do_inverse);
327  fft_ref_init(fft_nbits, do_inverse);
328  break;
329 #if CONFIG_FFT_FLOAT
330  case TRANSFORM_RDFT:
331  if (do_inverse)
332  av_log(NULL, AV_LOG_INFO,"IDFT_C2R");
333  else
334  av_log(NULL, AV_LOG_INFO,"DFT_R2C");
335  ff_rdft_init(r, fft_nbits, do_inverse ? IDFT_C2R : DFT_R2C);
336  fft_ref_init(fft_nbits, do_inverse);
337  break;
338 # if CONFIG_DCT
339  case TRANSFORM_DCT:
340  if (do_inverse)
341  av_log(NULL, AV_LOG_INFO,"DCT_III");
342  else
343  av_log(NULL, AV_LOG_INFO,"DCT_II");
344  ff_dct_init(d, fft_nbits, do_inverse ? DCT_III : DCT_II);
345  break;
346 # endif
347 #endif
348  default:
349  av_log(NULL, AV_LOG_ERROR, "Requested transform not supported\n");
350  return 1;
351  }
352  av_log(NULL, AV_LOG_INFO," %d test\n", fft_size);
353 
354  /* generate random data */
355 
356  for (i = 0; i < fft_size; i++) {
357  tab1[i].re = frandom(&prng);
358  tab1[i].im = frandom(&prng);
359  }
360 
361  /* checking result */
362  av_log(NULL, AV_LOG_INFO,"Checking...\n");
363 
364  switch (transform) {
365  case TRANSFORM_MDCT:
366  if (do_inverse) {
367  imdct_ref((FFTSample *)tab_ref, (FFTSample *)tab1, fft_nbits);
368  m->imdct_calc(m, tab2, (FFTSample *)tab1);
369  err = check_diff((FFTSample *)tab_ref, tab2, fft_size, scale);
370  } else {
371  mdct_ref((FFTSample *)tab_ref, (FFTSample *)tab1, fft_nbits);
372 
373  m->mdct_calc(m, tab2, (FFTSample *)tab1);
374 
375  err = check_diff((FFTSample *)tab_ref, tab2, fft_size / 2, scale);
376  }
377  break;
378  case TRANSFORM_FFT:
379  memcpy(tab, tab1, fft_size * sizeof(FFTComplex));
380  s->fft_permute(s, tab);
381  s->fft_calc(s, tab);
382 
383  fft_ref(tab_ref, tab1, fft_nbits);
384  err = check_diff((FFTSample *)tab_ref, (FFTSample *)tab, fft_size * 2, 1.0);
385  break;
386 #if CONFIG_FFT_FLOAT
387  case TRANSFORM_RDFT:
388  fft_size_2 = fft_size >> 1;
389  if (do_inverse) {
390  tab1[ 0].im = 0;
391  tab1[fft_size_2].im = 0;
392  for (i = 1; i < fft_size_2; i++) {
393  tab1[fft_size_2+i].re = tab1[fft_size_2-i].re;
394  tab1[fft_size_2+i].im = -tab1[fft_size_2-i].im;
395  }
396 
397  memcpy(tab2, tab1, fft_size * sizeof(FFTSample));
398  tab2[1] = tab1[fft_size_2].re;
399 
400  r->rdft_calc(r, tab2);
401  fft_ref(tab_ref, tab1, fft_nbits);
402  for (i = 0; i < fft_size; i++) {
403  tab[i].re = tab2[i];
404  tab[i].im = 0;
405  }
406  err = check_diff((float *)tab_ref, (float *)tab, fft_size * 2, 0.5);
407  } else {
408  for (i = 0; i < fft_size; i++) {
409  tab2[i] = tab1[i].re;
410  tab1[i].im = 0;
411  }
412  r->rdft_calc(r, tab2);
413  fft_ref(tab_ref, tab1, fft_nbits);
414  tab_ref[0].im = tab_ref[fft_size_2].re;
415  err = check_diff((float *)tab_ref, (float *)tab2, fft_size, 1.0);
416  }
417  break;
418  case TRANSFORM_DCT:
419  memcpy(tab, tab1, fft_size * sizeof(FFTComplex));
420  d->dct_calc(d, (FFTSample *)tab);
421  if (do_inverse) {
422  idct_ref((FFTSample*)tab_ref, (FFTSample *)tab1, fft_nbits);
423  } else {
424  dct_ref((FFTSample*)tab_ref, (FFTSample *)tab1, fft_nbits);
425  }
426  err = check_diff((float *)tab_ref, (float *)tab, fft_size, 1.0);
427  break;
428 #endif
429  }
430 
431  /* do a speed test */
432 
433  if (do_speed) {
434  int64_t time_start, duration;
435  int nb_its;
436 
437  av_log(NULL, AV_LOG_INFO,"Speed test...\n");
438  /* we measure during about 1 seconds */
439  nb_its = 1;
440  for(;;) {
441  time_start = av_gettime();
442  for (it = 0; it < nb_its; it++) {
443  switch (transform) {
444  case TRANSFORM_MDCT:
445  if (do_inverse) {
446  m->imdct_calc(m, (FFTSample *)tab, (FFTSample *)tab1);
447  } else {
448  m->mdct_calc(m, (FFTSample *)tab, (FFTSample *)tab1);
449  }
450  break;
451  case TRANSFORM_FFT:
452  memcpy(tab, tab1, fft_size * sizeof(FFTComplex));
453  s->fft_calc(s, tab);
454  break;
455 #if CONFIG_FFT_FLOAT
456  case TRANSFORM_RDFT:
457  memcpy(tab2, tab1, fft_size * sizeof(FFTSample));
458  r->rdft_calc(r, tab2);
459  break;
460  case TRANSFORM_DCT:
461  memcpy(tab2, tab1, fft_size * sizeof(FFTSample));
462  d->dct_calc(d, tab2);
463  break;
464 #endif
465  }
466  }
467  duration = av_gettime() - time_start;
468  if (duration >= 1000000)
469  break;
470  nb_its *= 2;
471  }
472  av_log(NULL, AV_LOG_INFO,"time: %0.1f us/transform [total time=%0.2f s its=%d]\n",
473  (double)duration / nb_its,
474  (double)duration / 1000000.0,
475  nb_its);
476  }
477 
478  switch (transform) {
479  case TRANSFORM_MDCT:
480  ff_mdct_end(m);
481  break;
482  case TRANSFORM_FFT:
483  ff_fft_end(s);
484  break;
485 #if CONFIG_FFT_FLOAT
486  case TRANSFORM_RDFT:
487  ff_rdft_end(r);
488  break;
489 # if CONFIG_DCT
490  case TRANSFORM_DCT:
491  ff_dct_end(d);
492  break;
493 # endif
494 #endif
495  }
496 
497  av_free(tab);
498  av_free(tab1);
499  av_free(tab2);
500  av_free(tab_ref);
501  av_free(exptab);
502 
503  if (err)
504  printf("Error: %d.\n", err);
505 
506  return !!err;
507 }