falkTX
/
FFmpeg
mirror of https://github.com/falkTX/FFmpeg.git
1
0
Fork 0
Code Issues 0 Releases 338 Wiki Activity
You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
40183 Commits
32 Branches
759MB
Tree: 12eeced8b4
Branches Tags
${ item.name }
Create branch ${ searchTerm }
from '12eeced8b4'
${ noResults }
FFmpeg/libavcodec/dct-test.c

589 lines
16KB
Raw Blame History 

  1. /*

  2.  * (c) 2001 Fabrice Bellard

  3.  *     2007 Marc Hoffman <marc.hoffman@analog.com>

  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.  * DCT test (c) 2001 Fabrice Bellard

  25.  * Started from sample code by Juan J. Sierralta P.

  26.  */

  27. 

  28. #include <stdlib.h>

  29. #include <stdio.h>

  30. #include <string.h>

  31. #include <sys/time.h>

  32. #include <unistd.h>

  33. #include <math.h>

  34. 

  35. #include "libavutil/cpu.h"

  36. #include "libavutil/common.h"

  37. #include "libavutil/lfg.h"

  38. 

  39. #include "simple_idct.h"

  40. #include "aandcttab.h"

  41. #include "faandct.h"

  42. #include "faanidct.h"

  43. #include "x86/idct_xvid.h"

  44. #include "dctref.h"

  45. 

  46. #undef printf

  47. 

  48. void ff_mmx_idct(DCTELEM *data);

  49. void ff_mmxext_idct(DCTELEM *data);

  50. 

  51. // BFIN

  52. void ff_bfin_idct(DCTELEM *block);

  53. void ff_bfin_fdct(DCTELEM *block);

  54. 

  55. // ALTIVEC

  56. void ff_fdct_altivec(DCTELEM *block);

  57. //void ff_idct_altivec(DCTELEM *block);?? no routine

  58. 

  59. // ARM

  60. void ff_j_rev_dct_arm(DCTELEM *data);

  61. void ff_simple_idct_arm(DCTELEM *data);

  62. void ff_simple_idct_armv5te(DCTELEM *data);

  63. void ff_simple_idct_armv6(DCTELEM *data);

  64. void ff_simple_idct_neon(DCTELEM *data);

  65. 

  66. void ff_simple_idct_axp(DCTELEM *data);

  67. 

  68. struct algo {

  69.     const char *name;

  70.     void (*func)(DCTELEM *block);

  71.     enum formattag { NO_PERM, MMX_PERM, MMX_SIMPLE_PERM, SCALE_PERM,

  72.                      SSE2_PERM, PARTTRANS_PERM, TRANSPOSE_PERM } format;

  73.     int mm_support;

  74.     int nonspec;

  75. };

  76. 

  77. static int cpu_flags;

  78. 

  79. static const struct algo fdct_tab[] = {

  80.     { "REF-DBL",        ff_ref_fdct,           NO_PERM    },

  81.     { "FAAN",           ff_faandct,            NO_PERM    },

  82.     { "IJG-AAN-INT",    ff_fdct_ifast,         SCALE_PERM },

  83.     { "IJG-LLM-INT",    ff_jpeg_fdct_islow_8,  NO_PERM    },

  84. 

  85. #if HAVE_MMX

  86.     { "MMX",            ff_fdct_mmx,           NO_PERM,   AV_CPU_FLAG_MMX     },

  87.     { "MMX2",           ff_fdct_mmx2,          NO_PERM,   AV_CPU_FLAG_MMX2    },

  88.     { "SSE2",           ff_fdct_sse2,          NO_PERM,   AV_CPU_FLAG_SSE2    },

  89. #endif

  90. 

  91. #if HAVE_ALTIVEC

  92.     { "altivecfdct",    ff_fdct_altivec,       NO_PERM,   AV_CPU_FLAG_ALTIVEC },

  93. #endif

  94. 

  95. #if ARCH_BFIN

  96.     { "BFINfdct",       ff_bfin_fdct,          NO_PERM  },

  97. #endif

  98. 

  99.     { 0 }

  100. };

  101. 

  102. #if HAVE_MMX && HAVE_YASM

  103. void ff_prores_idct_put_10_sse2(uint16_t *dst, int linesize,

  104.                                 DCTELEM *block, int16_t *qmat);

  105. 

  106. static void ff_prores_idct_put_10_sse2_wrap(DCTELEM *dst){

  107.     int16_t qmat[64]; int i;

  108.     int16_t tmp[64];

  109. 

  110.     for(i=0; i<64; i++){

  111.         qmat[i]=4;

  112.         tmp[i]= dst[i];

  113.     }

  114.     ff_prores_idct_put_10_sse2(dst, 16, tmp, qmat);

  115. }

  116. #endif

  117. 

  118. static const struct algo idct_tab[] = {

  119.     { "FAANI",          ff_faanidct,           NO_PERM  },

  120.     { "REF-DBL",        ff_ref_idct,           NO_PERM  },

  121.     { "INT",            ff_j_rev_dct,          MMX_PERM },

  122.     { "SIMPLE-C",       ff_simple_idct_8,      NO_PERM  },

  123. 

  124. #if HAVE_MMX

  125. #if CONFIG_GPL

  126.     { "LIBMPEG2-MMX",   ff_mmx_idct,           MMX_PERM,  AV_CPU_FLAG_MMX,  1 },

  127.     { "LIBMPEG2-MMX2",  ff_mmxext_idct,        MMX_PERM,  AV_CPU_FLAG_MMX2, 1 },

  128. #endif

  129.     { "SIMPLE-MMX",     ff_simple_idct_mmx,  MMX_SIMPLE_PERM, AV_CPU_FLAG_MMX },

  130.     { "XVID-MMX",       ff_idct_xvid_mmx,      NO_PERM,   AV_CPU_FLAG_MMX,  1 },

  131.     { "XVID-MMX2",      ff_idct_xvid_mmx2,     NO_PERM,   AV_CPU_FLAG_MMX2, 1 },

  132.     { "XVID-SSE2",      ff_idct_xvid_sse2,     SSE2_PERM, AV_CPU_FLAG_SSE2, 1 },

  133. #if ARCH_X86_64 && HAVE_YASM

  134.     { "PR-SSE2",        ff_prores_idct_put_10_sse2_wrap,     TRANSPOSE_PERM, AV_CPU_FLAG_SSE2, 1 },

  135. #endif

  136. #endif

  137. 

  138. #if ARCH_BFIN

  139.     { "BFINidct",       ff_bfin_idct,          NO_PERM  },

  140. #endif

  141. 

  142. #if ARCH_ARM

  143.     { "SIMPLE-ARM",     ff_simple_idct_arm,    NO_PERM  },

  144.     { "INT-ARM",        ff_j_rev_dct_arm,      MMX_PERM },

  145. #endif

  146. #if HAVE_ARMV5TE

  147.     { "SIMPLE-ARMV5TE", ff_simple_idct_armv5te,NO_PERM  },

  148. #endif

  149. #if HAVE_ARMV6

  150.     { "SIMPLE-ARMV6",   ff_simple_idct_armv6,  MMX_PERM },

  151. #endif

  152. #if HAVE_NEON

  153.     { "SIMPLE-NEON",    ff_simple_idct_neon,   PARTTRANS_PERM },

  154. #endif

  155. 

  156. #if ARCH_ALPHA

  157.     { "SIMPLE-ALPHA",   ff_simple_idct_axp,    NO_PERM },

  158. #endif

  159. 

  160.     { 0 }

  161. };

  162. 

  163. #define AANSCALE_BITS 12

  164. 

  165. static int64_t gettime(void)

  166. {

  167.     struct timeval tv;

  168.     gettimeofday(&tv, NULL);

  169.     return (int64_t)tv.tv_sec * 1000000 + tv.tv_usec;

  170. }

  171. 

  172. #define NB_ITS 20000

  173. #define NB_ITS_SPEED 50000

  174. 

  175. static short idct_mmx_perm[64];

  176. 

  177. static short idct_simple_mmx_perm[64] = {

  178.     0x00, 0x08, 0x04, 0x09, 0x01, 0x0C, 0x05, 0x0D,

  179.     0x10, 0x18, 0x14, 0x19, 0x11, 0x1C, 0x15, 0x1D,

  180.     0x20, 0x28, 0x24, 0x29, 0x21, 0x2C, 0x25, 0x2D,

  181.     0x12, 0x1A, 0x16, 0x1B, 0x13, 0x1E, 0x17, 0x1F,

  182.     0x02, 0x0A, 0x06, 0x0B, 0x03, 0x0E, 0x07, 0x0F,

  183.     0x30, 0x38, 0x34, 0x39, 0x31, 0x3C, 0x35, 0x3D,

  184.     0x22, 0x2A, 0x26, 0x2B, 0x23, 0x2E, 0x27, 0x2F,

  185.     0x32, 0x3A, 0x36, 0x3B, 0x33, 0x3E, 0x37, 0x3F,

  186. };

  187. 

  188. static const uint8_t idct_sse2_row_perm[8] = { 0, 4, 1, 5, 2, 6, 3, 7 };

  189. 

  190. static void idct_mmx_init(void)

  191. {

  192.     int i;

  193. 

  194.     /* the mmx/mmxext idct uses a reordered input, so we patch scan tables */

  195.     for (i = 0; i < 64; i++) {

  196.         idct_mmx_perm[i] = (i & 0x38) | ((i & 6) >> 1) | ((i & 1) << 2);

  197.     }

  198. }

  199. 

  200. DECLARE_ALIGNED(16, static DCTELEM, block)[64];

  201. DECLARE_ALIGNED(8,  static DCTELEM, block1)[64];

  202. 

  203. static inline void mmx_emms(void)

  204. {

  205. #if HAVE_MMX

  206.     if (cpu_flags & AV_CPU_FLAG_MMX)

  207.         __asm__ volatile ("emms\n\t");

  208. #endif

  209. }

  210. 

  211. static void init_block(DCTELEM block[64], int test, int is_idct, AVLFG *prng, int vals)

  212. {

  213.     int i, j;

  214. 

  215.     memset(block, 0, 64 * sizeof(*block));

  216. 

  217.     switch (test) {

  218.     case 0:

  219.         for (i = 0; i < 64; i++)

  220.             block[i] = (av_lfg_get(prng) % (2*vals)) -vals;

  221.         if (is_idct) {

  222.             ff_ref_fdct(block);

  223.             for (i = 0; i < 64; i++)

  224.                 block[i] >>= 3;

  225.         }

  226.         break;

  227.     case 1:

  228.         j = av_lfg_get(prng) % 10 + 1;

  229.         for (i = 0; i < j; i++)

  230.             block[av_lfg_get(prng) % 64] = av_lfg_get(prng) % (2*vals) -vals;

  231.         break;

  232.     case 2:

  233.         block[ 0] = av_lfg_get(prng) % (16*vals) - (8*vals);

  234.         block[63] = (block[0] & 1) ^ 1;

  235.         break;

  236.     }

  237. }

  238. 

  239. static void permute(DCTELEM dst[64], const DCTELEM src[64], int perm)

  240. {

  241.     int i;

  242. 

  243.     if (perm == MMX_PERM) {

  244.         for (i = 0; i < 64; i++)

  245.             dst[idct_mmx_perm[i]] = src[i];

  246.     } else if (perm == MMX_SIMPLE_PERM) {

  247.         for (i = 0; i < 64; i++)

  248.             dst[idct_simple_mmx_perm[i]] = src[i];

  249.     } else if (perm == SSE2_PERM) {

  250.         for (i = 0; i < 64; i++)

  251.             dst[(i & 0x38) | idct_sse2_row_perm[i & 7]] = src[i];

  252.     } else if (perm == PARTTRANS_PERM) {

  253.         for (i = 0; i < 64; i++)

  254.             dst[(i & 0x24) | ((i & 3) << 3) | ((i >> 3) & 3)] = src[i];

  255.     } else if (perm == TRANSPOSE_PERM) {

  256.         for (i = 0; i < 64; i++)

  257.             dst[(i>>3) | ((i<<3)&0x38)] = src[i];

  258.     } else {

  259.         for (i = 0; i < 64; i++)

  260.             dst[i] = src[i];

  261.     }

  262. }

  263. 

  264. static int dct_error(const struct algo *dct, int test, int is_idct, int speed, const int bits)

  265. {

  266.     void (*ref)(DCTELEM *block) = is_idct ? ff_ref_idct : ff_ref_fdct;

  267.     int it, i, scale;

  268.     int err_inf, v;

  269.     int64_t err2, ti, ti1, it1, err_sum = 0;

  270.     int64_t sysErr[64], sysErrMax = 0;

  271.     int maxout = 0;

  272.     int blockSumErrMax = 0, blockSumErr;

  273.     AVLFG prng;

  274.     const int vals=1<<bits;

  275.     double omse, ome;

  276.     int spec_err;

  277. 

  278.     av_lfg_init(&prng, 1);

  279. 

  280.     err_inf = 0;

  281.     err2 = 0;

  282.     for (i = 0; i < 64; i++)

  283.         sysErr[i] = 0;

  284.     for (it = 0; it < NB_ITS; it++) {

  285.         init_block(block1, test, is_idct, &prng, vals);

  286.         permute(block, block1, dct->format);

  287. 

  288.         dct->func(block);

  289.         mmx_emms();

  290. 

  291.         if (dct->format == SCALE_PERM) {

  292.             for (i = 0; i < 64; i++) {

  293.                 scale = 8 * (1 << (AANSCALE_BITS + 11)) / ff_aanscales[i];

  294.                 block[i] = (block[i] * scale) >> AANSCALE_BITS;

  295.             }

  296.         }

  297. 

  298.         ref(block1);

  299. 

  300.         blockSumErr = 0;

  301.         for (i = 0; i < 64; i++) {

  302.             int err = block[i] - block1[i];

  303.             err_sum += err;

  304.             v = abs(err);

  305.             if (v > err_inf)

  306.                 err_inf = v;

  307.             err2 += v * v;

  308.             sysErr[i] += block[i] - block1[i];

  309.             blockSumErr += v;

  310.             if (abs(block[i]) > maxout)

  311.                 maxout = abs(block[i]);

  312.         }

  313.         if (blockSumErrMax < blockSumErr)

  314.             blockSumErrMax = blockSumErr;

  315.     }

  316.     for (i = 0; i < 64; i++)

  317.         sysErrMax = FFMAX(sysErrMax, FFABS(sysErr[i]));

  318. 

  319.     for (i = 0; i < 64; i++) {

  320.         if (i % 8 == 0)

  321.             printf("\n");

  322.         printf("%7d ", (int) sysErr[i]);

  323.     }

  324.     printf("\n");

  325. 

  326.     omse = (double) err2 / NB_ITS / 64;

  327.     ome  = (double) err_sum / NB_ITS / 64;

  328. 

  329.     spec_err = is_idct && (err_inf > 1 || omse > 0.02 || fabs(ome) > 0.0015);

  330. 

  331.     printf("%s %s: max_err=%d omse=%0.8f ome=%0.8f syserr=%0.8f maxout=%d blockSumErr=%d\n",

  332.            is_idct ? "IDCT" : "DCT", dct->name, err_inf,

  333.            omse, ome, (double) sysErrMax / NB_ITS,

  334.            maxout, blockSumErrMax);

  335. 

  336.     if (spec_err && !dct->nonspec)

  337.         return 1;

  338. 

  339.     if (!speed)

  340.         return 0;

  341. 

  342.     /* speed test */

  343. 

  344.     init_block(block, test, is_idct, &prng, vals);

  345.     permute(block1, block, dct->format);

  346. 

  347.     ti = gettime();

  348.     it1 = 0;

  349.     do {

  350.         for (it = 0; it < NB_ITS_SPEED; it++) {

  351.             memcpy(block, block1, sizeof(block));

  352.             dct->func(block);

  353.         }

  354.         it1 += NB_ITS_SPEED;

  355.         ti1 = gettime() - ti;

  356.     } while (ti1 < 1000000);

  357.     mmx_emms();

  358. 

  359.     printf("%s %s: %0.1f kdct/s\n", is_idct ? "IDCT" : "DCT", dct->name,

  360.            (double) it1 * 1000.0 / (double) ti1);

  361. 

  362.     return 0;

  363. }

  364. 

  365. DECLARE_ALIGNED(8, static uint8_t, img_dest)[64];

  366. DECLARE_ALIGNED(8, static uint8_t, img_dest1)[64];

  367. 

  368. static void idct248_ref(uint8_t *dest, int linesize, int16_t *block)

  369. {

  370.     static int init;

  371.     static double c8[8][8];

  372.     static double c4[4][4];

  373.     double block1[64], block2[64], block3[64];

  374.     double s, sum, v;

  375.     int i, j, k;

  376. 

  377.     if (!init) {

  378.         init = 1;

  379. 

  380.         for (i = 0; i < 8; i++) {

  381.             sum = 0;

  382.             for (j = 0; j < 8; j++) {

  383.                 s = (i == 0) ? sqrt(1.0 / 8.0) : sqrt(1.0 / 4.0);

  384.                 c8[i][j] = s * cos(M_PI * i * (j + 0.5) / 8.0);

  385.                 sum += c8[i][j] * c8[i][j];

  386.             }

  387.         }

  388. 

  389.         for (i = 0; i < 4; i++) {

  390.             sum = 0;

  391.             for (j = 0; j < 4; j++) {

  392.                 s = (i == 0) ? sqrt(1.0 / 4.0) : sqrt(1.0 / 2.0);

  393.                 c4[i][j] = s * cos(M_PI * i * (j + 0.5) / 4.0);

  394.                 sum += c4[i][j] * c4[i][j];

  395.             }

  396.         }

  397.     }

  398. 

  399.     /* butterfly */

  400.     s = 0.5 * sqrt(2.0);

  401.     for (i = 0; i < 4; i++) {

  402.         for (j = 0; j < 8; j++) {

  403.             block1[8 * (2 * i) + j] =

  404.                 (block[8 * (2 * i) + j] + block[8 * (2 * i + 1) + j]) * s;

  405.             block1[8 * (2 * i + 1) + j] =

  406.                 (block[8 * (2 * i) + j] - block[8 * (2 * i + 1) + j]) * s;

  407.         }

  408.     }

  409. 

  410.     /* idct8 on lines */

  411.     for (i = 0; i < 8; i++) {

  412.         for (j = 0; j < 8; j++) {

  413.             sum = 0;

  414.             for (k = 0; k < 8; k++)

  415.                 sum += c8[k][j] * block1[8 * i + k];

  416.             block2[8 * i + j] = sum;

  417.         }

  418.     }

  419. 

  420.     /* idct4 */

  421.     for (i = 0; i < 8; i++) {

  422.         for (j = 0; j < 4; j++) {

  423.             /* top */

  424.             sum = 0;

  425.             for (k = 0; k < 4; k++)

  426.                 sum += c4[k][j] * block2[8 * (2 * k) + i];

  427.             block3[8 * (2 * j) + i] = sum;

  428. 

  429.             /* bottom */

  430.             sum = 0;

  431.             for (k = 0; k < 4; k++)

  432.                 sum += c4[k][j] * block2[8 * (2 * k + 1) + i];

  433.             block3[8 * (2 * j + 1) + i] = sum;

  434.         }

  435.     }

  436. 

  437.     /* clamp and store the result */

  438.     for (i = 0; i < 8; i++) {

  439.         for (j = 0; j < 8; j++) {

  440.             v = block3[8 * i + j];

  441.             if      (v < 0)   v = 0;

  442.             else if (v > 255) v = 255;

  443.             dest[i * linesize + j] = (int) rint(v);

  444.         }

  445.     }

  446. }

  447. 

  448. static void idct248_error(const char *name,

  449.                           void (*idct248_put)(uint8_t *dest, int line_size,

  450.                                               int16_t *block),

  451.                           int speed)

  452. {

  453.     int it, i, it1, ti, ti1, err_max, v;

  454.     AVLFG prng;

  455. 

  456.     av_lfg_init(&prng, 1);

  457. 

  458.     /* just one test to see if code is correct (precision is less

  459.        important here) */

  460.     err_max = 0;

  461.     for (it = 0; it < NB_ITS; it++) {

  462.         /* XXX: use forward transform to generate values */

  463.         for (i = 0; i < 64; i++)

  464.             block1[i] = av_lfg_get(&prng) % 256 - 128;

  465.         block1[0] += 1024;

  466. 

  467.         for (i = 0; i < 64; i++)

  468.             block[i] = block1[i];

  469.         idct248_ref(img_dest1, 8, block);

  470. 

  471.         for (i = 0; i < 64; i++)

  472.             block[i] = block1[i];

  473.         idct248_put(img_dest, 8, block);

  474. 

  475.         for (i = 0; i < 64; i++) {

  476.             v = abs((int) img_dest[i] - (int) img_dest1[i]);

  477.             if (v == 255)

  478.                 printf("%d %d\n", img_dest[i], img_dest1[i]);

  479.             if (v > err_max)

  480.                 err_max = v;

  481.         }

  482. #if 0

  483.         printf("ref=\n");

  484.         for(i=0;i<8;i++) {

  485.             int j;

  486.             for(j=0;j<8;j++) {

  487.                 printf(" %3d", img_dest1[i*8+j]);

  488.             }

  489.             printf("\n");

  490.         }

  491. 

  492.         printf("out=\n");

  493.         for(i=0;i<8;i++) {

  494.             int j;

  495.             for(j=0;j<8;j++) {

  496.                 printf(" %3d", img_dest[i*8+j]);

  497.             }

  498.             printf("\n");

  499.         }

  500. #endif

  501.     }

  502.     printf("%s %s: err_inf=%d\n", 1 ? "IDCT248" : "DCT248", name, err_max);

  503. 

  504.     if (!speed)

  505.         return;

  506. 

  507.     ti = gettime();

  508.     it1 = 0;

  509.     do {

  510.         for (it = 0; it < NB_ITS_SPEED; it++) {

  511.             for (i = 0; i < 64; i++)

  512.                 block[i] = block1[i];

  513.             idct248_put(img_dest, 8, block);

  514.         }

  515.         it1 += NB_ITS_SPEED;

  516.         ti1 = gettime() - ti;

  517.     } while (ti1 < 1000000);

  518.     mmx_emms();

  519. 

  520.     printf("%s %s: %0.1f kdct/s\n", 1 ? "IDCT248" : "DCT248", name,

  521.            (double) it1 * 1000.0 / (double) ti1);

  522. }

  523. 

  524. static void help(void)

  525. {

  526.     printf("dct-test [-i] [<test-number>] [<bits>]\n"

  527.            "test-number 0 -> test with random matrixes\n"

  528.            "            1 -> test with random sparse matrixes\n"

  529.            "            2 -> do 3. test from mpeg4 std\n"

  530.            "bits        Number of time domain bits to use, 8 is default\n"

  531.            "-i          test IDCT implementations\n"

  532.            "-4          test IDCT248 implementations\n"

  533.            "-t          speed test\n");

  534. }

  535. 

  536. int main(int argc, char **argv)

  537. {

  538.     int test_idct = 0, test_248_dct = 0;

  539.     int c, i;

  540.     int test = 1;

  541.     int speed = 0;

  542.     int err = 0;

  543.     int bits=8;

  544. 

  545.     cpu_flags = av_get_cpu_flags();

  546. 

  547.     ff_ref_dct_init();

  548.     idct_mmx_init();

  549. 

  550.     for (;;) {

  551.         c = getopt(argc, argv, "ih4t");

  552.         if (c == -1)

  553.             break;

  554.         switch (c) {

  555.         case 'i':

  556.             test_idct = 1;

  557.             break;

  558.         case '4':

  559.             test_248_dct = 1;

  560.             break;

  561.         case 't':

  562.             speed = 1;

  563.             break;

  564.         default:

  565.         case 'h':

  566.             help();

  567.             return 0;

  568.         }

  569.     }

  570. 

  571.     if (optind < argc)

  572.         test = atoi(argv[optind]);

  573.     if(optind+1 < argc) bits= atoi(argv[optind+1]);

  574. 

  575.     printf("ffmpeg DCT/IDCT test\n");

  576. 

  577.     if (test_248_dct) {

  578.         idct248_error("SIMPLE-C", ff_simple_idct248_put, speed);

  579.     } else {

  580.         const struct algo *algos = test_idct ? idct_tab : fdct_tab;

  581.         for (i = 0; algos[i].name; i++)

  582.             if (!(~cpu_flags & algos[i].mm_support)) {

  583.                 err |= dct_error(&algos[i], test, test_idct, speed, bits);

  584.             }

  585.     }

  586. 

  587.     return err;

  588. }