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.
59781 Commits
32 Branches
303MB
Tree: 2ce4543286
Branches Tags
${ item.name }
Create branch ${ searchTerm }
from '2ce4543286'
${ noResults }
FFmpeg/libavcodec/dct-test.c

597 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 "config.h"

  29. #include <stdlib.h>

  30. #include <stdio.h>

  31. #include <string.h>

  32. #if HAVE_UNISTD_H

  33. #include <unistd.h>

  34. #endif

  35. #include <math.h>

  36. 

  37. #include "libavutil/cpu.h"

  38. #include "libavutil/common.h"

  39. #include "libavutil/lfg.h"

  40. #include "libavutil/time.h"

  41. 

  42. #include "dct.h"

  43. #include "simple_idct.h"

  44. #include "aandcttab.h"

  45. #include "faandct.h"

  46. #include "faanidct.h"

  47. #include "x86/idct_xvid.h"

  48. #include "dctref.h"

  49. 

  50. // BFIN

  51. void ff_bfin_idct(int16_t *block);

  52. void ff_bfin_fdct(int16_t *block);

  53. 

  54. // ALTIVEC

  55. void ff_fdct_altivec(int16_t *block);

  56. 

  57. // ARM

  58. void ff_j_rev_dct_arm(int16_t *data);

  59. void ff_simple_idct_arm(int16_t *data);

  60. void ff_simple_idct_armv5te(int16_t *data);

  61. void ff_simple_idct_armv6(int16_t *data);

  62. void ff_simple_idct_neon(int16_t *data);

  63. 

  64. struct algo {

  65.     const char *name;

  66.     void (*func)(int16_t *block);

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

  68.                      SSE2_PERM, PARTTRANS_PERM, TRANSPOSE_PERM } format;

  69.     int mm_support;

  70.     int nonspec;

  71. };

  72. 

  73. static int cpu_flags;

  74. 

  75. static const struct algo fdct_tab[] = {

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

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

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

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

  80. 

  81. #if HAVE_MMX_INLINE

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

  83. #endif

  84. #if HAVE_MMXEXT_INLINE

  85.     { "MMXEXT",         ff_fdct_mmxext,        NO_PERM,   AV_CPU_FLAG_MMXEXT  },

  86. #endif

  87. #if HAVE_SSE2_INLINE

  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. static void ff_prores_idct_wrap(int16_t *dst){

  103.     DECLARE_ALIGNED(16, static int16_t, qmat)[64];

  104.     int i;

  105. 

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

  107.         qmat[i]=4;

  108.     }

  109.     ff_prores_idct(dst, qmat);

  110. }

  111. #if ARCH_X86_64 && HAVE_MMX && HAVE_YASM

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

  113.                                 int16_t *block, int16_t *qmat);

  114. 

  115. static void ff_prores_idct_put_10_sse2_wrap(int16_t *dst){

  116.     DECLARE_ALIGNED(16, static int16_t, qmat)[64];

  117.     DECLARE_ALIGNED(16, static int16_t, tmp)[64];

  118.     int i;

  119. 

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

  121.         qmat[i]=4;

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

  123.     }

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

  125. 

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

  127.          dst[i] -= 512;

  128.     }

  129. }

  130. #endif

  131. 

  132. static const struct algo idct_tab[] = {

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

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

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

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

  137.     { "PR-C",           ff_prores_idct_wrap,   NO_PERM, 0, 1 },

  138. 

  139. #if HAVE_MMX_INLINE

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

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

  142. #endif

  143. #if HAVE_MMXEXT_INLINE

  144.     { "XVID-MMXEXT",    ff_idct_xvid_mmxext,   NO_PERM,   AV_CPU_FLAG_MMXEXT, 1 },

  145. #endif

  146. #if HAVE_SSE2_INLINE

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

  148. #if ARCH_X86_64 && HAVE_YASM

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

  150. #endif

  151. #endif

  152. 

  153. #if ARCH_BFIN

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

  155. #endif

  156. 

  157. #if ARCH_ARM

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

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

  160. #endif

  161. #if HAVE_ARMV5TE

  162.     { "SIMPLE-ARMV5TE", ff_simple_idct_armv5te,NO_PERM,   AV_CPU_FLAG_ARMV5TE },

  163. #endif

  164. #if HAVE_ARMV6

  165.     { "SIMPLE-ARMV6",   ff_simple_idct_armv6,  MMX_PERM,  AV_CPU_FLAG_ARMV6   },

  166. #endif

  167. #if HAVE_NEON

  168.     { "SIMPLE-NEON",    ff_simple_idct_neon, PARTTRANS_PERM, AV_CPU_FLAG_NEON },

  169. #endif

  170. 

  171.     { 0 }

  172. };

  173. 

  174. #define AANSCALE_BITS 12

  175. 

  176. #define NB_ITS 20000

  177. #define NB_ITS_SPEED 50000

  178. 

  179. static short idct_mmx_perm[64];

  180. 

  181. static short idct_simple_mmx_perm[64] = {

  182.     0x00, 0x08, 0x04, 0x09, 0x01, 0x0C, 0x05, 0x0D,

  183.     0x10, 0x18, 0x14, 0x19, 0x11, 0x1C, 0x15, 0x1D,

  184.     0x20, 0x28, 0x24, 0x29, 0x21, 0x2C, 0x25, 0x2D,

  185.     0x12, 0x1A, 0x16, 0x1B, 0x13, 0x1E, 0x17, 0x1F,

  186.     0x02, 0x0A, 0x06, 0x0B, 0x03, 0x0E, 0x07, 0x0F,

  187.     0x30, 0x38, 0x34, 0x39, 0x31, 0x3C, 0x35, 0x3D,

  188.     0x22, 0x2A, 0x26, 0x2B, 0x23, 0x2E, 0x27, 0x2F,

  189.     0x32, 0x3A, 0x36, 0x3B, 0x33, 0x3E, 0x37, 0x3F,

  190. };

  191. 

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

  193. 

  194. static void idct_mmx_init(void)

  195. {

  196.     int i;

  197. 

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

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

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

  201.     }

  202. }

  203. 

  204. DECLARE_ALIGNED(16, static int16_t, block)[64];

  205. DECLARE_ALIGNED(8,  static int16_t, block1)[64];

  206. 

  207. static void init_block(int16_t block[64], int test, int is_idct, AVLFG *prng, int vals)

  208. {

  209.     int i, j;

  210. 

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

  212. 

  213.     switch (test) {

  214.     case 0:

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

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

  217.         if (is_idct) {

  218.             ff_ref_fdct(block);

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

  220.                 block[i] >>= 3;

  221.         }

  222.         break;

  223.     case 1:

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

  225.         for (i = 0; i < j; i++) {

  226.             int idx = av_lfg_get(prng) % 64;

  227.             block[idx] = av_lfg_get(prng) % (2*vals) -vals;

  228.         }

  229.         break;

  230.     case 2:

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

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

  233.         break;

  234.     }

  235. }

  236. 

  237. static void permute(int16_t dst[64], const int16_t src[64], int perm)

  238. {

  239.     int i;

  240. 

  241.     if (perm == MMX_PERM) {

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

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

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

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

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

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

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

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

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

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

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

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

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

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

  256.     } else {

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

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

  259.     }

  260. }

  261. 

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

  263. {

  264.     void (*ref)(int16_t *block) = is_idct ? ff_ref_idct : ff_ref_fdct;

  265.     int it, i, scale;

  266.     int err_inf, v;

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

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

  269.     int maxout = 0;

  270.     int blockSumErrMax = 0, blockSumErr;

  271.     AVLFG prng;

  272.     const int vals=1<<bits;

  273.     double omse, ome;

  274.     int spec_err;

  275. 

  276.     av_lfg_init(&prng, 1);

  277. 

  278.     err_inf = 0;

  279.     err2 = 0;

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

  281.         sysErr[i] = 0;

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

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

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

  285. 

  286.         dct->func(block);

  287.         emms_c();

  288. 

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

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

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

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

  293.             }

  294.         }

  295. 

  296.         ref(block1);

  297.         if (!strcmp(dct->name, "PR-SSE2"))

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

  299.                 block1[i] = av_clip(block1[i], 4-512, 1019-512);

  300. 

  301.         blockSumErr = 0;

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

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

  304.             err_sum += err;

  305.             v = abs(err);

  306.             if (v > err_inf)

  307.                 err_inf = v;

  308.             err2 += v * v;

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

  310.             blockSumErr += v;

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

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

  313.         }

  314.         if (blockSumErrMax < blockSumErr)

  315.             blockSumErrMax = blockSumErr;

  316.     }

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

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

  319. 

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

  321.         if (i % 8 == 0)

  322.             printf("\n");

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

  324.     }

  325.     printf("\n");

  326. 

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

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

  329. 

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

  331. 

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

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

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

  335.            maxout, blockSumErrMax);

  336. 

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

  338.         return 1;

  339. 

  340.     if (!speed)

  341.         return 0;

  342. 

  343.     /* speed test */

  344. 

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

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

  347. 

  348.     ti = av_gettime();

  349.     it1 = 0;

  350.     do {

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

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

  353.             dct->func(block);

  354.         }

  355.         emms_c();

  356.         it1 += NB_ITS_SPEED;

  357.         ti1 = av_gettime() - ti;

  358.     } while (ti1 < 1000000);

  359. 

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

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

  362. 

  363.     return 0;

  364. }

  365. 

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

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

  368. 

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

  370. {

  371.     static int init;

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

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

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

  375.     double s, sum, v;

  376.     int i, j, k;

  377. 

  378.     if (!init) {

  379.         init = 1;

  380. 

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

  382.             sum = 0;

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

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

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

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

  387.             }

  388.         }

  389. 

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

  391.             sum = 0;

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

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

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

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

  396.             }

  397.         }

  398.     }

  399. 

  400.     /* butterfly */

  401.     s = 0.5 * sqrt(2.0);

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

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

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

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

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

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

  408.         }

  409.     }

  410. 

  411.     /* idct8 on lines */

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

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

  414.             sum = 0;

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

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

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

  418.         }

  419.     }

  420. 

  421.     /* idct4 */

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

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

  424.             /* top */

  425.             sum = 0;

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

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

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

  429. 

  430.             /* bottom */

  431.             sum = 0;

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

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

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

  435.         }

  436.     }

  437. 

  438.     /* clamp and store the result */

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

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

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

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

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

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

  445.         }

  446.     }

  447. }

  448. 

  449. static void idct248_error(const char *name,

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

  451.                                               int16_t *block),

  452.                           int speed)

  453. {

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

  455.     AVLFG prng;

  456. 

  457.     av_lfg_init(&prng, 1);

  458. 

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

  460.        important here) */

  461.     err_max = 0;

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

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

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

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

  466.         block1[0] += 1024;

  467. 

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

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

  470.         idct248_ref(img_dest1, 8, block);

  471. 

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

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

  474.         idct248_put(img_dest, 8, block);

  475. 

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

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

  478.             if (v == 255)

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

  480.             if (v > err_max)

  481.                 err_max = v;

  482.         }

  483. #if 0

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

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

  486.             int j;

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

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

  489.             }

  490.             printf("\n");

  491.         }

  492. 

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

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

  495.             int j;

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

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

  498.             }

  499.             printf("\n");

  500.         }

  501. #endif

  502.     }

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

  504. 

  505.     if (!speed)

  506.         return;

  507. 

  508.     ti = av_gettime();

  509.     it1 = 0;

  510.     do {

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

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

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

  514.             idct248_put(img_dest, 8, block);

  515.         }

  516.         emms_c();

  517.         it1 += NB_ITS_SPEED;

  518.         ti1 = av_gettime() - ti;

  519.     } while (ti1 < 1000000);

  520. 

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

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

  523. }

  524. 

  525. static void help(void)

  526. {

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

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

  529.            "            1 -> test with random sparse matrixes\n"

  530.            "            2 -> do 3. test from mpeg4 std\n"

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

  532.            "-i          test IDCT implementations\n"

  533.            "-4          test IDCT248 implementations\n"

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

  535. }

  536. 

  537. #if !HAVE_GETOPT

  538. #include "compat/getopt.c"

  539. #endif

  540. 

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

  542. {

  543.     int test_idct = 0, test_248_dct = 0;

  544.     int c, i;

  545.     int test = 1;

  546.     int speed = 0;

  547.     int err = 0;

  548.     int bits=8;

  549. 

  550.     cpu_flags = av_get_cpu_flags();

  551. 

  552.     ff_ref_dct_init();

  553.     idct_mmx_init();

  554. 

  555.     for (;;) {

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

  557.         if (c == -1)

  558.             break;

  559.         switch (c) {

  560.         case 'i':

  561.             test_idct = 1;

  562.             break;

  563.         case '4':

  564.             test_248_dct = 1;

  565.             break;

  566.         case 't':

  567.             speed = 1;

  568.             break;

  569.         default:

  570.         case 'h':

  571.             help();

  572.             return 0;

  573.         }

  574.     }

  575. 

  576.     if (optind < argc)

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

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

  579. 

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

  581. 

  582.     if (test_248_dct) {

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

  584.     } else {

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

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

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

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

  589.             }

  590.     }

  591. 

  592.     if (err)

  593.         printf("Error: %d.\n", err);

  594. 

  595.     return !!err;

  596. }