Diff of /trunk/xvidcore/examples/xvid_bench.c

-revision 851, Sat Feb 15 15:22:19 2003 UTC
+revision 2021, Wed Jul  6 10:25:14 2011 UTC
 Line 1
- /**************************************************************************
+ /*****************************************************************************
   *
-  *      XVID MPEG-4 VIDEO CODEC - Unit tests and benches
+  *  XVID MPEG-4 VIDEO CODEC
+  *  - Unit tests and benches -
+  *
+  *  Copyright(C) 2002 Pascal Massimino <skal@planet-d.net>
   *
   *      This program is free software; you can redistribute it and/or modify
   *      it under the terms of the GNU General Public License as published by
-Line 14
+Line 17
   *
   *      You should have received a copy of the GNU General Public License
   *      along with this program; if not, write to the Free Software
-  *      Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
+  *  Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307 USA
+  *
+  * $Id$
   *
-  *************************************************************************/
+  ****************************************************************************/
- /************************************************************************
+ /*****************************************************************************
   *
   *  'Reference' output is at the end of file.
-  *  Don't take the checksums and crc too seriouly, they aren't
-  *  bullet-proof (should plug some .md5 here)...
   *
   *   compiles with something like:
   *   gcc -o xvid_bench xvid_bench.c  -I../src/ -lxvidcore -lm
   *
-  *      History:
+  ****************************************************************************/
-  *
-  *      06.06.2002  initial coding      -Skal-
-  *
-  *************************************************************************/
  #include <stdio.h>
  #include <stdlib.h>
- #include <sys/time.h>  // for gettimeofday
+ #include <string.h>    /* for memset */
- #include <string.h>    // for memset
  #include <assert.h>
+ #ifndef WIN32
+ #include <sys/time.h>   /* for gettimeofday */
+ #else
+ #include <time.h>
+ #endif
  #include "xvid.h"
  // inner guts
+ #include "portab.h"
  #include "dct/idct.h"
  #include "dct/fdct.h"
  #include "image/colorspace.h"
  #include "image/interpolate8x8.h"
  #include "utils/mem_transfer.h"
- #include "quant/quant_h263.h"
+ #include "quant/quant.h"
- #include "quant/quant_mpeg4.h"
  #include "motion/sad.h"
  #include "utils/emms.h"
  #include "utils/timer.h"
  #include "quant/quant_matrix.c"
  #include "bitstream/cbp.h"
+ #include "bitstream/bitstream.h"
  #include <math.h>
- const int speed_ref = 100;  // on slow machines, decrease this value
+ #ifndef M_PI
+ #define M_PI            3.14159265358979323846
+ #endif
+ int speed_ref = 100;  /* on slow machines, decrease this value */
+ int verbose = 0;
+ unsigned int cpu_mask;
  /*********************************************************************
   * misc
-Line 65
+Line 78
   /* returns time in micro-s*/
  double gettime_usec()
  {
+ #ifndef WIN32
    struct timeval  tv;
    gettimeofday(&tv, 0);
    return tv.tv_sec*1.0e6 + tv.tv_usec;
+ #else
+         clock_t clk;
+         clk = clock();
+         return clk * 1000. / CLOCKS_PER_SEC;  /* clock() returns time in Milliseconds */
+ #endif
  }
   /* returns squared deviates (mean(v*v)-mean(v)^2) of a 8x8 block */
-Line 95
+Line 114
    unsigned int cpu;
  } CPU;
- CPU cpu_list[] =
+ CPU cpu_list[] = {
- { { "PLAINC", 0 }
+         { "PLAINC ", 0 },
- , { "MMX   ", XVID_CPU_MMX }
+ #if defined(ARCH_IS_IA32) || defined(ARCH_IS_X86_64)
- , { "MMXEXT", XVID_CPU_MMXEXT | XVID_CPU_MMX }
+         { "MMX    ", XVID_CPU_MMX },
- , { "SSE2  ", XVID_CPU_SSE2 | XVID_CPU_MMX }
+         { "MMXEXT ", XVID_CPU_MMXEXT | XVID_CPU_MMX },
- , { "3DNOW ", XVID_CPU_3DNOW }
+         { "SSE2   ", XVID_CPU_SSE2 | XVID_CPU_MMX },
- , { "3DNOWE", XVID_CPU_3DNOWEXT }
+         { "SSE3   ", XVID_CPU_SSE3 | XVID_CPU_SSE2 | XVID_CPU_MMX },
- , { "IA64  ", XVID_CPU_IA64 }
+         { "SSE41  ", XVID_CPU_SSE41| XVID_CPU_SSE3 | XVID_CPU_SSE2 | XVID_CPU_MMX },
- //, { "TSC   ", XVID_CPU_TSC }
+         { "3DNOW  ", XVID_CPU_3DNOW },
- , { 0, 0 } }
+         { "3DNOWE ", XVID_CPU_3DNOW | XVID_CPU_3DNOWEXT },
+ #endif
- , cpu_short_list[] =
+ #ifdef ARCH_IS_PPC
- { { "PLAINC", 0 }
+         { "ALTIVEC", XVID_CPU_ALTIVEC },
- , { "MMX   ", XVID_CPU_MMX }
+ #endif
- //, { "MMXEXT", XVID_CPU_MMXEXT | XVID_CPU_MMX }
+ #ifdef ARCH_IS_IA64
- , { "IA64  ", XVID_CPU_IA64 }
+ //      { "IA64   ", XVID_CPU_IA64 },
- , { 0, 0 } }
+ #endif
+ //      { "TSC    ", XVID_CPU_TSC },
- , cpu_short_list2[] =
+         { 0, 0 }
- { { "PLAINC", 0 }
+ };
- , { "MMX   ", XVID_CPU_MMX }
- , { "SSE2  ", XVID_CPU_SSE2 | XVID_CPU_MMX }
- , { 0, 0 } };
  int init_cpu(CPU *cpu)
  {
-   int xerr, cpu_type;
+         xvid_gbl_info_t xinfo;
-   XVID_INIT_PARAM xinit;
+         /* Get the available CPU flags */
+         memset(&xinfo, 0, sizeof(xinfo));
+         xinfo.version = XVID_VERSION;
+         xvid_global(NULL, XVID_GBL_INFO, &xinfo, NULL);
-   cpu_type = check_cpu_features() & cpu->cpu;
+         /* Are we trying to test a subset of the host CPU features */
-   xinit.cpu_flags = cpu_type | XVID_CPU_FORCE;
+         if ((xinfo.cpu_flags & cpu->cpu) == cpu->cpu) {
-   //    xinit.cpu_flags = XVID_CPU_MMX | XVID_CPU_FORCE;
+                 int xerr;
-   xerr = xvid_init(NULL, 0, &xinit, NULL);
+                 xvid_gbl_init_t xinit;
-   if (cpu->cpu>0 && (cpu_type==0 || xerr!=XVID_ERR_OK)) {
+                 memset(&xinit, 0, sizeof(xinit));
-     printf( "%s - skipped...\n", cpu->name );
+                 xinit.cpu_flags = cpu->cpu | XVID_CPU_FORCE;
+                 xinit.version = XVID_VERSION;
+                 xerr = xvid_global(NULL, XVID_GBL_INIT, &xinit, NULL);
+                 if (xerr==XVID_ERR_FAIL) {
+                         /* libxvidcore failed to init */
      return 0;
    }
+         } else {
+                 /* The host CPU doesn't support some required feature for this test */
+                 return(0);
+         }
    return 1;
  }
+ #define CRC32_REMAINDER 0xCBF43926
+ #define CRC32_INITIAL 0xffffffff
+ #define DO1(c, crc) ((crc) = crc32tab[((unsigned int)((crc)>>24) ^ (*c++)) & 0xff] ^ ((crc) << 8))
+ #define DO2(c, crc)  DO1(c, crc); DO1(c, crc);
+ #define DO4(c, crc)  DO2(c, crc); DO2(c, crc);
+ #define DO8(c, crc)  DO4(c, crc); DO4(c, crc);
+ /******************************************************************************
+ * Precomputed AAL5 CRC32 lookup table
+ ******************************************************************************/
+ static unsigned long crc32tab[256] = {
+x00000000L, 0x04C11DB7L, 0x09823B6EL, 0x0D4326D9L,
+x130476DCL, 0x17C56B6BL, 0x1A864DB2L, 0x1E475005L,
+x2608EDB8L, 0x22C9F00FL, 0x2F8AD6D6L, 0x2B4BCB61L,
+x350C9B64L, 0x31CD86D3L, 0x3C8EA00AL, 0x384FBDBDL,
+x4C11DB70L, 0x48D0C6C7L, 0x4593E01EL, 0x4152FDA9L,
+x5F15ADACL, 0x5BD4B01BL, 0x569796C2L, 0x52568B75L,
+x6A1936C8L, 0x6ED82B7FL, 0x639B0DA6L, 0x675A1011L,
+x791D4014L, 0x7DDC5DA3L, 0x709F7B7AL, 0x745E66CDL,
+x9823B6E0L, 0x9CE2AB57L, 0x91A18D8EL, 0x95609039L,
+x8B27C03CL, 0x8FE6DD8BL, 0x82A5FB52L, 0x8664E6E5L,
+xBE2B5B58L, 0xBAEA46EFL, 0xB7A96036L, 0xB3687D81L,
+xAD2F2D84L, 0xA9EE3033L, 0xA4AD16EAL, 0xA06C0B5DL,
+xD4326D90L, 0xD0F37027L, 0xDDB056FEL, 0xD9714B49L,
+xC7361B4CL, 0xC3F706FBL, 0xCEB42022L, 0xCA753D95L,
+xF23A8028L, 0xF6FB9D9FL, 0xFBB8BB46L, 0xFF79A6F1L,
+xE13EF6F4L, 0xE5FFEB43L, 0xE8BCCD9AL, 0xEC7DD02DL,
+x34867077L, 0x30476DC0L, 0x3D044B19L, 0x39C556AEL,
+x278206ABL, 0x23431B1CL, 0x2E003DC5L, 0x2AC12072L,
+x128E9DCFL, 0x164F8078L, 0x1B0CA6A1L, 0x1FCDBB16L,
+x018AEB13L, 0x054BF6A4L, 0x0808D07DL, 0x0CC9CDCAL,
+x7897AB07L, 0x7C56B6B0L, 0x71159069L, 0x75D48DDEL,
+x6B93DDDBL, 0x6F52C06CL, 0x6211E6B5L, 0x66D0FB02L,
+x5E9F46BFL, 0x5A5E5B08L, 0x571D7DD1L, 0x53DC6066L,
+x4D9B3063L, 0x495A2DD4L, 0x44190B0DL, 0x40D816BAL,
+xACA5C697L, 0xA864DB20L, 0xA527FDF9L, 0xA1E6E04EL,
+xBFA1B04BL, 0xBB60ADFCL, 0xB6238B25L, 0xB2E29692L,
+x8AAD2B2FL, 0x8E6C3698L, 0x832F1041L, 0x87EE0DF6L,
+x99A95DF3L, 0x9D684044L, 0x902B669DL, 0x94EA7B2AL,
+xE0B41DE7L, 0xE4750050L, 0xE9362689L, 0xEDF73B3EL,
+xF3B06B3BL, 0xF771768CL, 0xFA325055L, 0xFEF34DE2L,
+xC6BCF05FL, 0xC27DEDE8L, 0xCF3ECB31L, 0xCBFFD686L,
+xD5B88683L, 0xD1799B34L, 0xDC3ABDEDL, 0xD8FBA05AL,
+x690CE0EEL, 0x6DCDFD59L, 0x608EDB80L, 0x644FC637L,
+x7A089632L, 0x7EC98B85L, 0x738AAD5CL, 0x774BB0EBL,
+x4F040D56L, 0x4BC510E1L, 0x46863638L, 0x42472B8FL,
+x5C007B8AL, 0x58C1663DL, 0x558240E4L, 0x51435D53L,
+x251D3B9EL, 0x21DC2629L, 0x2C9F00F0L, 0x285E1D47L,
+x36194D42L, 0x32D850F5L, 0x3F9B762CL, 0x3B5A6B9BL,
+x0315D626L, 0x07D4CB91L, 0x0A97ED48L, 0x0E56F0FFL,
+x1011A0FAL, 0x14D0BD4DL, 0x19939B94L, 0x1D528623L,
+xF12F560EL, 0xF5EE4BB9L, 0xF8AD6D60L, 0xFC6C70D7L,
+xE22B20D2L, 0xE6EA3D65L, 0xEBA91BBCL, 0xEF68060BL,
+xD727BBB6L, 0xD3E6A601L, 0xDEA580D8L, 0xDA649D6FL,
+xC423CD6AL, 0xC0E2D0DDL, 0xCDA1F604L, 0xC960EBB3L,
+xBD3E8D7EL, 0xB9FF90C9L, 0xB4BCB610L, 0xB07DABA7L,
+xAE3AFBA2L, 0xAAFBE615L, 0xA7B8C0CCL, 0xA379DD7BL,
+x9B3660C6L, 0x9FF77D71L, 0x92B45BA8L, 0x9675461FL,
+x8832161AL, 0x8CF30BADL, 0x81B02D74L, 0x857130C3L,
+x5D8A9099L, 0x594B8D2EL, 0x5408ABF7L, 0x50C9B640L,
+x4E8EE645L, 0x4A4FFBF2L, 0x470CDD2BL, 0x43CDC09CL,
+x7B827D21L, 0x7F436096L, 0x7200464FL, 0x76C15BF8L,
+x68860BFDL, 0x6C47164AL, 0x61043093L, 0x65C52D24L,
+x119B4BE9L, 0x155A565EL, 0x18197087L, 0x1CD86D30L,
+x029F3D35L, 0x065E2082L, 0x0B1D065BL, 0x0FDC1BECL,
+x3793A651L, 0x3352BBE6L, 0x3E119D3FL, 0x3AD08088L,
+x2497D08DL, 0x2056CD3AL, 0x2D15EBE3L, 0x29D4F654L,
+xC5A92679L, 0xC1683BCEL, 0xCC2B1D17L, 0xC8EA00A0L,
+xD6AD50A5L, 0xD26C4D12L, 0xDF2F6BCBL, 0xDBEE767CL,
+xE3A1CBC1L, 0xE760D676L, 0xEA23F0AFL, 0xEEE2ED18L,
+xF0A5BD1DL, 0xF464A0AAL, 0xF9278673L, 0xFDE69BC4L,
+x89B8FD09L, 0x8D79E0BEL, 0x803AC667L, 0x84FBDBD0L,
+x9ABC8BD5L, 0x9E7D9662L, 0x933EB0BBL, 0x97FFAD0CL,
+xAFB010B1L, 0xAB710D06L, 0xA6322BDFL, 0xA2F33668L,
+xBCB4666DL, 0xB8757BDAL, 0xB5365D03L, 0xB1F740B4L
+ };
+ uint32_t
+ calc_crc(uint8_t *mem, int len, uint32_t crc)
+ {
+         while( len >= 8) {
+                 DO8(mem, crc);
+                 len -= 8;
+         }
+         while( len ) {
+                 DO1(mem, crc);
+                 len--;
+         }
+         return crc;
+ }
+ void byte_swap(uint8_t *mem, int len, int element_size) {
+ #ifdef ARCH_IS_BIG_ENDIAN
+         int i;
+         if(element_size == 1) {
+                 /* No need to swap */
+         } else if(element_size == 2) {
+                 uint8_t temp[2];
+                 for(i=0; i < (len/2); i++ ) {
+                         temp[0] = mem[0];
+                         temp[1] = mem[1];
+                         mem[0] = temp[1];
+                         mem[1] = temp[0];
+                         mem += 2;
+                 }
+         } else if(element_size == 4) {
+                 uint8_t temp[4];
+                 for(i=0; i < (len/4); i++ ) {
+                         temp[0] = mem[0];
+                         temp[1] = mem[1];
+                         temp[2] = mem[2];
+                         temp[3] = mem[3];
+                         mem[0] = temp[3];
+                         mem[1] = temp[2];
+                         mem[2] = temp[1];
+                         mem[3] = temp[0];
+                         mem += 4;
+                 }
+         } else {
+                 printf("ERROR: byte_swap unsupported element_size(%u)\n", element_size);
+         }
+ #endif
+ }
  /*********************************************************************
   * test DCT
   *********************************************************************/
-Line 148
+Line 311
    int tst;
    CPU *cpu;
    int i;
-   short iDst0[8*8], iDst[8*8], fDst[8*8];
+         DECLARE_ALIGNED_MATRIX(iDst0, 8, 8, short, 16);
+         DECLARE_ALIGNED_MATRIX(iDst,  8, 8, short, 16);
+         DECLARE_ALIGNED_MATRIX(fDst,  8, 8, short, 16);
    double overhead;
    printf( "\n ===== test fdct/idct =====\n" );
-Line 186
+Line 351
        MSE += delta*delta;
      }
      PSNR = (MSE==0.) ? 1.e6 : -4.3429448*log( MSE/64. );
-     printf( "%s -  %.3f usec       PSNR=%.3f  MSE=%.3f\n",
+                 printf( "%s -  %.3f usec       PSNR=%.3f  MSE=%.3f %s\n",
-       cpu->name, t, PSNR, MSE );
+                                 cpu->name, t, PSNR, MSE,
-     if (ABS(MSE)>=64) printf( "*** CRC ERROR! ***\n" );
+                                 (ABS(MSE)>=64)? "| ERROR" :"");
    }
  }
-Line 202
+Line 367
    int tst;
    CPU *cpu;
    int i;
-   uint8_t Cur[16*16], Ref1[16*16], Ref2[16*16];
+         DECLARE_ALIGNED_MATRIX(Cur,  16, 16, uint8_t, 16);
+         DECLARE_ALIGNED_MATRIX(Ref1, 16, 16, uint8_t, 16);
+         DECLARE_ALIGNED_MATRIX(Ref2, 16, 16, uint8_t, 16);
    printf( "\n ======  test SAD ======\n" );
    for(i=0; i<16*16;++i) {
-Line 223
+Line 390
      for(tst=0; tst<nb_tests; ++tst) s = sad8(Cur, Ref1, 16);
      emms();
      t = (gettime_usec() - t) / nb_tests;
-     printf( "%s - sad8    %.3f usec       sad=%d\n", cpu->name, t, s );
+                 printf("%s - sad8    %.3f usec       sad=%d %s\n",
-     if (s!=3776) printf( "*** CRC ERROR! ***\n" );
+                            cpu->name, t, s,
+                            (s!=3776)?"| ERROR": "" );
      t = gettime_usec();
      emms();
      for(tst=0; tst<nb_tests; ++tst) s = sad16(Cur, Ref1, 16, -1);
      emms();
      t = (gettime_usec() - t) / nb_tests;
-     printf( "%s - sad16   %.3f usec       sad=%d\n", cpu->name, t, s );
+                 printf("%s - sad16   %.3f usec       sad=%d %s\n",
-     if (s!=27214) printf( "*** CRC ERROR! ***\n" );
+                            cpu->name, t, s,
+                            (s!=27214)?"| ERROR": "" );
      t = gettime_usec();
      emms();
      for(tst=0; tst<nb_tests; ++tst) s = sad16bi(Cur, Ref1, Ref2, 16);
      emms();
      t = (gettime_usec() - t) / nb_tests;
-     printf( "%s - sad16bi %.3f usec       sad=%d\n", cpu->name, t, s );
+                 printf( "%s - sad16bi %.3f usec       sad=%d %s\n",
-     if (s!=26274) printf( "*** CRC ERROR! ***\n" );
+                                 cpu->name, t, s,
+                                 (s!=26274)?"| ERROR": "" );
+                 t = gettime_usec();
+                 emms();
+                 for(tst=0; tst<nb_tests; ++tst) s = sad8bi(Cur, Ref1, Ref2, 8);
+                 emms();
+                 t = (gettime_usec() - t) / nb_tests;
+                 printf( "%s - sad8bi %.3f usec       sad=%d %s\n",
+                                 cpu->name, t, s,
+                                 (s!=4002)?"| ERROR": "" );
      t = gettime_usec();
      emms();
      for(tst=0; tst<nb_tests; ++tst) s = dev16(Cur, 16);
      emms();
      t = (gettime_usec() - t) / nb_tests;
-     printf( "%s - dev16   %.3f usec       sad=%d\n", cpu->name, t, s );
+                 printf( "%s - dev16   %.3f usec       sad=%d %s\n",
-     if (s!=3344) printf( "*** CRC ERROR! ***\n" );
+                                 cpu->name, t, s,
+                                 (s!=3344)?"| ERROR": "" );
      printf( " --- \n" );
    }
-Line 266
+Line 446
  #define LEAVE \
      emms();                             \
      t = (gettime_usec() - t) / nb_tests;  \
-     iCrc = 0;                           \
+         iCrc = calc_crc((uint8_t*)Dst, sizeof(Dst), CRC32_INITIAL)
-     for(i=0; i<16*8; ++i) { iCrc += Dst[i]^i; }
  #define TEST_MB(FUNC, R)                \
      ENTER                               \
-Line 285
+Line 464
    const int nb_tests = 2000*speed_ref;
    CPU *cpu;
    const uint8_t Src0[16*9] = {
-         // try to have every possible combinaison of rounding...
+                 /* try to have every possible combinaison of rounding... */
-, 0, 1, 0, 2, 0, 3, 0, 4             ,0,0,0, 0,0,0,0
+, 0, 1, 0, 2, 0, 3, 0, 4             ,0,0,0, 0,0,0,0,
-     , 0, 1, 1, 1, 2, 1, 3, 1, 3             ,0,0,0, 0,0,0,0
+, 1, 1, 1, 2, 1, 3, 1, 3             ,0,0,0, 0,0,0,0,
-     , 0, 2, 1, 2, 2, 2, 3, 2, 2             ,0,0,0, 0,0,0,0
+, 2, 1, 2, 2, 2, 3, 2, 2             ,0,0,0, 0,0,0,0,
-     , 0, 3, 1, 3, 2, 3, 3, 3, 1             ,0,0,0, 0,0,0,0
+, 3, 1, 3, 2, 3, 3, 3, 1             ,0,0,0, 0,0,0,0,
-     , 1, 3, 0, 2, 1, 0, 2, 3, 4             ,0,0,0, 0,0,0,0
+, 3, 0, 2, 1, 0, 2, 3, 4             ,0,0,0, 0,0,0,0,
-     , 2, 2, 1, 2, 0, 1, 3, 5, 3             ,0,0,0, 0,0,0,0
+, 2, 1, 2, 0, 1, 3, 5, 3             ,0,0,0, 0,0,0,0,
-     , 3, 1, 2, 3, 1, 2, 2, 6, 2             ,0,0,0, 0,0,0,0
+, 1, 2, 3, 1, 2, 2, 6, 2             ,0,0,0, 0,0,0,0,
-     , 1, 0, 1, 3, 0, 3, 1, 6, 1             ,0,0,0, 0,0,0,0
+, 0, 1, 3, 0, 3, 1, 6, 1             ,0,0,0, 0,0,0,0,
-     , 4, 3, 2, 1, 2, 3, 4, 0, 3             ,0,0,0, 0,0,0,0
+, 3, 2, 1, 2, 3, 4, 0, 3             ,0,0,0, 0,0,0,0
    };
    uint8_t Dst[16*8] = {0};
-Line 309
+Line 488
        continue;
      TEST_MB(interpolate8x8_halfpel_h, 0);
-     printf( "%s - interp- h-round0 %.3f usec       iCrc=%d\n", cpu->name, t, iCrc );
+                 printf("%s - interp- h-round0 %.3f usec       crc32=0x%08x %s\n",
-     if (iCrc!=8107) printf( "*** CRC ERROR! ***\n" );
+                            cpu->name, t, iCrc,
+                            (iCrc!=0x115381ba)?"| ERROR": "" );
      TEST_MB(interpolate8x8_halfpel_h, 1);
-     printf( "%s -           round1 %.3f usec       iCrc=%d\n", cpu->name, t, iCrc );
+                 printf("%s -           round1 %.3f usec       crc32=0x%08x %s\n",
-     if (iCrc!=8100) printf( "*** CRC ERROR! ***\n" );
+                            cpu->name, t, iCrc,
+                            (iCrc!=0x2b1f528f)?"| ERROR": "" );
      TEST_MB(interpolate8x8_halfpel_v, 0);
-     printf( "%s - interp- v-round0 %.3f usec       iCrc=%d\n", cpu->name, t, iCrc );
+                 printf("%s - interp- v-round0 %.3f usec       crc32=0x%08x %s\n",
-     if (iCrc!=8108) printf( "*** CRC ERROR! ***\n" );
+                            cpu->name, t, iCrc,
+                            (iCrc!=0x423cdcc7)?"| ERROR": "" );
      TEST_MB(interpolate8x8_halfpel_v, 1);
-     printf( "%s -           round1 %.3f usec       iCrc=%d\n", cpu->name, t, iCrc );
+                 printf("%s -           round1 %.3f usec       crc32=0x%08x %s\n",
-     if (iCrc!=8105) printf( "*** CRC ERROR! ***\n" );
+                            cpu->name, t, iCrc,
+                            (iCrc!=0x42202efe)?"| ERROR": "" );
      TEST_MB(interpolate8x8_halfpel_hv, 0);
-     printf( "%s - interp-hv-round0 %.3f usec       iCrc=%d\n", cpu->name, t, iCrc );
+                 printf("%s - interp-hv-round0 %.3f usec       crc32=0x%08x %s\n",
-     if (iCrc!=8112) printf( "*** CRC ERROR! ***\n" );
+                            cpu->name, t, iCrc,
+                            (iCrc!=0xd198d387)?"| ERROR": "" );
      TEST_MB(interpolate8x8_halfpel_hv, 1);
-     printf( "%s -           round1 %.3f usec       iCrc=%d\n", cpu->name, t, iCrc );
+                 printf("%s -           round1 %.3f usec       crc32=0x%08x %s\n",
-     if (iCrc!=8103) printf( "*** CRC ERROR! ***\n" );
+                            cpu->name, t, iCrc,
+                            (iCrc!=0x9ecfd921)?"| ERROR": "" );
-        // this is a new function, as of 06.06.2002
+                 /* this is a new function, as of 06.06.2002 */
  #if 0
      TEST_MB2(interpolate8x8_avrg);
-     printf( "%s - interpolate8x8_c %.3f usec       iCrc=%d\n", cpu->name, t, iCrc );
+                 printf("%s - interpolate8x8_c %.3f usec       crc32=0x%08x %s\n",
-     if (iCrc!=8107) printf( "*** CRC ERROR! ***\n" );
+                            cpu->name, t, iCrc,
+                            (iCrc!=8107)?"| ERROR": "" );
  #endif
+     /* New functions for field prediction by CK 1.10.2005 */
+ #pragma NEW8X4
+                 TEST_MB(interpolate8x4_halfpel_h, 0);
+                 printf("%s - interpfield-h -round0 %.3f usec       crc32=0x%08x %s\n",
+                            cpu->name, t, iCrc,
+                            (iCrc!=0x9538d6df)?"| ERROR": "" );
+                 TEST_MB(interpolate8x4_halfpel_h, 1);
+                 printf("%s -                round1 %.3f usec       crc32=0x%08x %s\n",
+                            cpu->name, t, iCrc,
+                            (iCrc!=0xde5f1db4)?"| ERROR": "" );
+                 TEST_MB(interpolate8x4_halfpel_v, 0);
+                 printf("%s - interpfield- v-round0 %.3f usec       crc32=0x%08x %s\n",
+                            cpu->name, t, iCrc,
+                            (iCrc!=0xea5a69ef)?"| ERROR": "" );
+                 TEST_MB(interpolate8x4_halfpel_v, 1);
+                 printf("%s -                round1 %.3f usec       crc32=0x%08x %s\n",
+                            cpu->name, t, iCrc,
+                            (iCrc!=0x4f10ec0f)?"| ERROR": "" );
+                 TEST_MB(interpolate8x4_halfpel_hv, 0);
+                 printf("%s - interpfield-hv-round0 %.3f usec       crc32=0x%08x %s\n",
+                            cpu->name, t, iCrc,
+                            (iCrc!=0xf97ee367)?"| ERROR": "" );
+                 TEST_MB(interpolate8x4_halfpel_hv, 1);
+                 printf("%s -                round1 %.3f usec       crc32=0x%08x %s\n",
+                            cpu->name, t, iCrc,
+                            (iCrc!=0xb6a9f581)?"| ERROR": "" );
+ /* End of 8x4 functions */
      printf( " --- \n" );
    }
  }
+ #undef ENTER
+ #undef LEAVE
+ #undef TEST_MB
+ #undef TEST_MB2
  /*********************************************************************
   * test transfer
   *********************************************************************/
-Line 375
+Line 601
      }                                         \
      emms();                                   \
      t = (gettime_usec()-t -overhead) / nb_tests;\
-     s = 0; for(i=0; i<8*32; ++i) { s += (DST)[i]^i; }
+ byte_swap((uint8_t*)(DST), 8*32*sizeof((DST)[0]), sizeof((DST)[0]));  \
+ s = calc_crc((uint8_t*)(DST), 8*32*sizeof((DST)[0]), CRC32_INITIAL)
  #define TEST_TRANSFER(FUNC, DST, SRC)         \
      TEST_TRANSFER_BEGIN(DST);                 \
-Line 401
+Line 628
      }                                         \
      emms();                                   \
      t = (gettime_usec()-t -overhead) / nb_tests;\
-     s = 0; for(i=0; i<8*32; ++i) { s += (DST)[i]; }
+ byte_swap((uint8_t*)(DST), 8*32*sizeof((DST)[0]), sizeof((DST)[0]));  \
+ s = calc_crc((uint8_t*)(DST), 8*32*sizeof((DST)[0]), CRC32_INITIAL)
  #define TEST_TRANSFER2(FUNC, DST, SRC, R1)    \
      TEST_TRANSFER2_BEGIN(DST,SRC);            \
-Line 418
+Line 646
    const int nb_tests = 4000*speed_ref;
    int i;
    CPU *cpu;
-   uint8_t  Src8[8*32], Dst8[8*32], Ref1[8*32], Ref2[8*32];
+ //      uint8_t  Src8[8*32], Dst8[8*32], Ref1[8*32], Ref2[8*32];
-   int16_t Src16[8*32], Dst16[8*32];
+ //      int16_t Src16[8*32], Dst16[8*32];
+   DECLARE_ALIGNED_MATRIX(Src8, 8, 32, uint8_t, CACHE_LINE);
+   DECLARE_ALIGNED_MATRIX(Dst8, 8, 32, uint8_t, CACHE_LINE);
+   DECLARE_ALIGNED_MATRIX(Ref1, 8, 32, uint8_t, CACHE_LINE);
+   DECLARE_ALIGNED_MATRIX(Ref2, 8, 32, uint8_t, CACHE_LINE);
+   DECLARE_ALIGNED_MATRIX(Src16, 8, 32, uint16_t, CACHE_LINE);
+   DECLARE_ALIGNED_MATRIX(Dst16, 8, 32, uint16_t, CACHE_LINE);
    printf( "\n ===  test transfer ===\n" );
-   for(cpu = cpu_short_list; cpu->name!=0; ++cpu)
+         for(cpu = cpu_list; cpu->name!=0; ++cpu)
    {
      double t, overhead;
      int tst, s;
-Line 432
+Line 666
        continue;
      TEST_TRANSFER(transfer_8to16copy, Dst16, Src8);
-     printf( "%s - 8to16     %.3f usec       crc=%d\n", cpu->name, t, s );
+                 printf("%s - 8to16     %.3f usec       crc32=0x%08x %s\n",
-     if (s!=28288) printf( "*** CRC ERROR! ***\n" );
+                            cpu->name, t, s,
+                            (s!=0x115814bb)?"| ERROR": "");
      TEST_TRANSFER(transfer_16to8copy, Dst8, Src16);
-     printf( "%s - 16to8     %.3f usec       crc=%d\n", cpu->name, t, s );
+                 printf( "%s - 16to8     %.3f usec       crc32=0x%08x %s\n",
-     if (s!=28288) printf( "*** CRC ERROR! ***\n" );
+                                 cpu->name, t, s,
+                                 (s!=0xee7ccbb4)?"| ERROR": "");
+     /* New functions for field prediction by CK 1.10.2005 */
+ #pragma NEW8X4
+                 TEST_TRANSFER(transfer8x4_copy, Dst8, Src8);
+                 printf("%s - 8to4      %.3f usec       crc32=0x%08x %s\n",
+                            cpu->name, t, s,
+                            (s!=0xbb9c3db5)?"| ERROR": "");
+ /* End of new functions */
      TEST_TRANSFER(transfer8x8_copy, Dst8, Src8);
-     printf( "%s - 8to8      %.3f usec       crc=%d\n", cpu->name, t, s );
+                 printf("%s - 8to8      %.3f usec       crc32=0x%08x %s\n",
-     if (s!=20352) printf( "*** CRC ERROR! ***\n" );
+                            cpu->name, t, s,
+                            (s!=0xd37b3295)?"| ERROR": "");
      TEST_TRANSFER(transfer_16to8add, Dst8, Src16);
-     printf( "%s - 16to8add  %.3f usec       crc=%d\n", cpu->name, t, s );
+                 printf("%s - 16to8add  %.3f usec       crc32=0x%08x %s\n",
-     if (s!=25536) printf( "*** CRC ERROR! ***\n" );
+                            cpu->name, t, s,
+                            (s!=0xdd817bf4)?"| ERROR": "" );
      TEST_TRANSFER2(transfer_8to16sub, Dst16, Src8, Ref1);
-     printf( "%s - 8to16sub  %.3f usec       crc1=%d ", cpu->name, t, s );
+                 {
-     if (s!=28064) printf( "*** CRC ERROR! ***\n" );
+                         int s1, s2;
-     s = 0; for(i=0; i<8*32; ++i) { s += (Src8[i]-Ref1[i])&i; }
+                         s1 = calc_crc((uint8_t*)Dst16, 8*32*sizeof(Dst16[0]), CRC32_INITIAL);
-     printf( "crc2=%d\n", s);
+                         s2 = calc_crc((uint8_t*)Src8, 8*32*sizeof(Src8[0]), CRC32_INITIAL);
-     if (s!=16256) printf( "*** CRC ERROR! ***\n" );
+                         printf("%s - 8to16sub  %.3f usec       crc32(1)=0x%08x crc32(2)=0x%08x %s %s\n",
- #if 1
+                                    cpu->name, t, s1, s2,
+                                    (s1!=0xa1e07163)?"| ERROR1": "",
+                                    (s2!=0xd86c5d23)?"| ERROR2": "" );
+                 }
      TEST_TRANSFER3(transfer_8to16sub2, Dst16, Src8, Ref1, Ref2);
-     printf( "%s - 8to16sub2 %.3f usec       crc=%d\n", cpu->name, t, s );
+                 printf("%s - 8to16sub2 %.3f usec       crc32=0x%08x %s\n",
-     if (s!=20384) printf( "*** CRC ERROR! ***\n" );
+                            cpu->name, t, s,
- //    for(i=0; i<64; ++i) printf( "[%d]", Dst16[i]);
+                            (s!=0x99b6c4c7)?"| ERROR": "" );
- //    printf("\n");
- #endif
      printf( " --- \n" );
    }
  }
-Line 470
+Line 719
  #define TEST_QUANT(FUNC, DST, SRC)              \
      t = gettime_usec();                         \
-     for(s=0,qm=1; qm<=255; ++qm) {              \
+ for(s=CRC32_INITIAL,qm=1; qm<=255; ++qm) {              \
        for(i=0; i<8*8; ++i) Quant[i] = qm;       \
-       set_inter_matrix( Quant );                \
+   set_inter_matrix( mpeg_quant_matrices, Quant );                \
        emms();                                   \
        for(q=1; q<=max_Q; ++q) {                 \
          for(tst=0; tst<nb_tests; ++tst)         \
-           (FUNC)((DST), (SRC), q);              \
+           (FUNC)((DST), (SRC), q, mpeg_quant_matrices);              \
-         for(i=0; i<64; ++i) s+=(DST)[i]^i^qm;   \
+         byte_swap((uint8_t*)(DST), 64*sizeof((DST)[0]), sizeof((DST)[0]));  \
+         s = calc_crc((uint8_t*)(DST), 64*sizeof((DST)[0]), s); \
        }                                         \
        emms();                                   \
      }                                           \
-     t = (gettime_usec()-t-overhead)/nb_tests/qm;\
+ t = (gettime_usec()-t-overhead)/nb_tests/qm
-     s = (s&0xffff)^(s>>16)
  #define TEST_QUANT2(FUNC, DST, SRC)             \
      t = gettime_usec();                         \
-     for(s=0,qm=1; qm<=255; ++qm) {              \
+ for(s=CRC32_INITIAL,qm=1; qm<=255; ++qm) {              \
        for(i=0; i<8*8; ++i) Quant[i] = qm;       \
-       set_intra_matrix( Quant );                \
+   set_intra_matrix( mpeg_quant_matrices, Quant );                \
        emms();                                   \
        for(q=1; q<=max_Q; ++q) {                 \
+         init_intra_matrix( mpeg_quant_matrices, q ); \
          for(tst=0; tst<nb_tests; ++tst)         \
-           (FUNC)((DST), (SRC), q, q);           \
+           (FUNC)((DST), (SRC), q, q, mpeg_quant_matrices);           \
-         for(i=0; i<64; ++i) s+=(DST)[i]^i^qm;   \
+         byte_swap((uint8_t*)(DST), 64*sizeof((DST)[0]), sizeof((DST)[0]));  \
+         s = calc_crc((uint8_t*)(DST), 64*sizeof((DST)[0]), s); \
        }                                         \
        emms();                                   \
      }                                           \
-     t = (gettime_usec()-t-overhead)/nb_tests/qm;\
+ t = (gettime_usec()-t-overhead)/nb_tests/qm
-     s = (s&0xffff)^(s>>16)
+ #define TEST_INTRA(REFFUNC, NEWFUNC, RANGE)              \
+ { int32_t i,q,s;\
+         DECLARE_ALIGNED_MATRIX(Src, 8, 8, int16_t, 16); \
+   DECLARE_ALIGNED_MATRIX(Dst, 8, 8, int16_t, 16); \
+   DECLARE_ALIGNED_MATRIX(Dst2,8, 8, int16_t, 16); \
+   for(q=1;q<=max_Q;q++)          \
+     for(s=-RANGE;s<RANGE;s++) { \
+       for(i=0;i<64;i++) Src[i]=s; \
+       (REFFUNC)((Dst),(Src),q,q,mpeg_quant_matrices);   \
+       (NEWFUNC)((Dst2),(Src),q,q,mpeg_quant_matrices);  \
+       for(i=0;i<64;i++)     \
+         if(Dst[i]!=Dst2[i]) printf("ERROR : " #NEWFUNC " i%d quant:%d input:%d C_result:%d ASM_result:%d\n",i,q,s,Dst[i],Dst2[i]);  \
+     }      \
+ }
+ #define TEST_INTER(REFFUNC, NEWFUNC, RANGE)              \
+ { int i,q,s;  \
+         DECLARE_ALIGNED_MATRIX(Src, 8, 8, int16_t, 16); \
+   DECLARE_ALIGNED_MATRIX(Dst, 8, 8, int16_t, 16); \
+   DECLARE_ALIGNED_MATRIX(Dst2,8, 8, int16_t, 16); \
+   for(q=1;q<=max_Q;q++)  \
+     for(s=-RANGE;s<RANGE;s++) {   \
+       for(i=0;i<64;i++) Src[i]=s; \
+       (REFFUNC)((Dst),(Src),q,mpeg_quant_matrices);  \
+       (NEWFUNC)((Dst2),(Src),q,mpeg_quant_matrices); \
+       emms();           \
+       for(i=0;i<64;i++) \
+         if(Dst[i]!=Dst2[i]) printf("ERROR : " #NEWFUNC " i%d quant:%d input:%d C_result:%d ASM_result:%d\n",i,q,s,Dst[i],Dst2[i]); \
+     } \
+ }
  void test_quant()
  {
-   const int nb_tests = 1*speed_ref;
+         const int32_t nb_tests = 1*speed_ref;
-   const int max_Q = 31;
+         const int32_t max_Q = 31;
-   int i, qm;
+         DECLARE_ALIGNED_MATRIX(mpeg_quant_matrices, 8, 64, uint16_t, 16);
+         int32_t i, qm;
    CPU *cpu;
-   int16_t  Src[8*8], Dst[8*8];
+         DECLARE_ALIGNED_MATRIX(Src, 8, 8, int16_t, 16);
+         DECLARE_ALIGNED_MATRIX(Dst, 8, 8, int16_t, 16);
+         DECLARE_ALIGNED_MATRIX(Dst2,8, 8, int16_t, 16);
    uint8_t Quant[8*8];
    printf( "\n =====  test quant =====\n" );
-     // we deliberately enfringe the norm's specified range [-127,127],
+ /* we deliberately enfringe the norm's specified range [-127,127], */
-     // to test the robustness of the iquant module
+ /* to test the robustness of the iquant module */
    for(i=0; i<64; ++i) {
      Src[i] = 1 + (i-32) * (i&6);
      Dst[i] = 0;
    }
-   for(cpu = cpu_short_list; cpu->name!=0; ++cpu)
+         for(cpu = cpu_list; cpu->name!=0; ++cpu)
    {
      double t, overhead;
-     int tst, q;
+                 int32_t tst, q;
      uint32_t s;
      if (!init_cpu(cpu))
        continue;
+                 // exhaustive tests to compare against the (ref) C-version
+                 TEST_INTRA(quant_h263_intra_c,   quant_h263_intra,    2048);
+                 TEST_INTRA(dequant_h263_intra_c, dequant_h263_intra , 512 );
+                 TEST_INTER(quant_h263_inter_c,   quant_h263_inter ,   2048);
+                 TEST_INTER(dequant_h263_inter_c, dequant_h263_inter , 512 );
      overhead = -gettime_usec();
      for(s=0,qm=1; qm<=255; ++qm) {
        for(i=0; i<8*8; ++i) Quant[i] = qm;
-       set_inter_matrix( Quant );
+                         set_inter_matrix(mpeg_quant_matrices, Quant );
        for(q=1; q<=max_Q; ++q)
          for(i=0; i<64; ++i) s+=Dst[i]^i^qm;
      }
      overhead += gettime_usec();
- #if 1
+                 TEST_QUANT2(quant_mpeg_intra, Dst, Src);
-     TEST_QUANT2(quant4_intra, Dst, Src);
+                 printf("%s -   quant_mpeg_intra %.3f usec       crc32=0x%08x %s\n",
-     printf( "%s -   quant4_intra %.3f usec       crc=%d\n", cpu->name, t, s );
+                            cpu->name, t, s,
-     if (s!=29809) printf( "*** CRC ERROR! ***\n" );
+                            (s!=0x3b999af6)? "| ERROR": "");
-     TEST_QUANT(quant4_inter, Dst, Src);
+                 TEST_QUANT(quant_mpeg_inter, Dst, Src);
-     printf( "%s -   quant4_inter %.3f usec       crc=%d\n", cpu->name, t, s );
+                 printf("%s -   quant_mpeg_inter %.3f usec       crc32=0x%08x %s\n",
-     if (s!=12574) printf( "*** CRC ERROR! ***\n" );
+                            cpu->name, t, s,
- #endif
+                            (s!=0xf6de7757)?"| ERROR": "");
- #if 1
-     TEST_QUANT2(dequant4_intra, Dst, Src);
+                 TEST_QUANT2(dequant_mpeg_intra, Dst, Src);
-     printf( "%s - dequant4_intra %.3f usec       crc=%d\n", cpu->name, t, s );
+                 printf("%s - dequant_mpeg_intra %.3f usec       crc32=0x%08x %s\n",
-     if (s!=24052) printf( "*** CRC ERROR! ***\n" );
+                            cpu->name, t, s,
+                            (s!=0x2def7bc7)?"| ERROR": "");
-     TEST_QUANT(dequant4_inter, Dst, Src);
-     printf( "%s - dequant4_inter %.3f usec       crc=%d\n", cpu->name, t, s );
+                 TEST_QUANT(dequant_mpeg_inter, Dst, Src);
-     if (s!=63847) printf( "*** CRC ERROR! ***\n" );
+                 printf("%s - dequant_mpeg_inter %.3f usec       crc32=0x%08x %s\n",
- #endif
+                            cpu->name, t, s,
- #if 1
+                            (s!=0xd878c722)?"| ERROR": "");
-     TEST_QUANT2(quant_intra, Dst, Src);
-     printf( "%s -    quant_intra %.3f usec       crc=%d\n", cpu->name, t, s );
+                 TEST_QUANT2(quant_h263_intra, Dst, Src);
-     if (s!=25662) printf( "*** CRC ERROR! ***\n" );
+                 printf("%s -   quant_h263_intra %.3f usec       crc32=0x%08x %s\n",
+                            cpu->name, t, s,
-     TEST_QUANT(quant_inter, Dst, Src);
+                            (s!=0x2eba9d43)?"| ERROR": "");
-     printf( "%s -    quant_inter %.3f usec       crc=%d\n", cpu->name, t, s );
-     if (s!=23972) printf( "*** CRC ERROR! ***\n" );
+                 TEST_QUANT(quant_h263_inter, Dst, Src);
- #endif
+                 printf("%s -   quant_h263_inter %.3f usec       crc32=0x%08x %s\n",
- #if 1
+                            cpu->name, t, s,
-     TEST_QUANT2(dequant_intra, Dst, Src);
+                            (s!=0xbd315a7e)?"| ERROR": "");
-     printf( "%s -  dequant_intra %.3f usec       crc=%d\n", cpu->name, t, s );
-     if (s!=49900) printf( "*** CRC ERROR! ***\n" );
+                 TEST_QUANT2(dequant_h263_intra, Dst, Src);
+                 printf("%s - dequant_h263_intra %.3f usec       crc32=0x%08x %s\n",
-     TEST_QUANT(dequant_inter, Dst, Src);
+                            cpu->name, t, s,
-     printf( "%s -  dequant_inter %.3f usec       crc=%d\n", cpu->name, t, s );
+                            (s!=0x9841212a)?"| ERROR": "");
-     if (s!=48899) printf( "*** CRC ERROR! ***\n" );
- #endif
+                 TEST_QUANT(dequant_h263_inter, Dst, Src);
+                 printf("%s - dequant_h263_inter %.3f usec       crc32=0x%08x %s\n",
+                            cpu->name, t, s,
+                            (s!=0xe7df8fba)?"| ERROR": "");
+                 printf( " --- \n" );
+         }
+ }
+ /*********************************************************************
+  * test distortion operators
+  *********************************************************************/
+ static void ieee_reseed(long s);
+ static long ieee_rand(int Min, int Max);
+ #define TEST_SSE(FUNCTION, SRC1, SRC2, STRIDE) \
+   do { \
+     t = gettime_usec(); \
+     tst = nb_tests; \
+     while((tst--)>0) sse = (FUNCTION)((SRC1), (SRC2), (STRIDE)); \
+     emms(); \
+     t = (gettime_usec() - t)/(double)nb_tests;  \
+   } while(0)
+ void test_sse()
+ {
+         const int nb_tests = 100000*speed_ref;
+         int i;
+         CPU *cpu;
+         DECLARE_ALIGNED_MATRIX(Src1, 8, 8, int16_t, 16);
+         DECLARE_ALIGNED_MATRIX(Src2, 8, 8, int16_t, 16);
+         DECLARE_ALIGNED_MATRIX(Src3, 8, 8, int16_t, 16);
+         DECLARE_ALIGNED_MATRIX(Src4, 8, 8, int16_t, 16);
+         printf( "\n =====  test sse =====\n" );
+         ieee_reseed(1);
+         for(i=0; i<64; ++i) {
+                 Src1[i] = ieee_rand(-2048, 2047);
+                 Src2[i] = ieee_rand(-2048, 2047);
+                 Src3[i] = ieee_rand(-2048, 2047);
+                 Src4[i] = ieee_rand(-2048, 2047);
+         }
+         for(cpu = cpu_list; cpu->name!=0; ++cpu)
+         {
+                 double t;
+                 int tst, sse;
+                 if (!init_cpu(cpu))
+                         continue;
+                 /* 16 bit element blocks */
+                 TEST_SSE(sse8_16bit, Src1, Src2, 16);
+                 printf("%s -   sse8_16bit#1 %.3f usec       sse=%d %s\n",
+                            cpu->name, t, sse, (sse!=182013834)?"| ERROR": "");
+                 TEST_SSE(sse8_16bit, Src1, Src3, 16);
+                 printf("%s -   sse8_16bit#2 %.3f usec       sse=%d %s\n",
+                            cpu->name, t, sse, (sse!=142545203)?"| ERROR": "");
+                 TEST_SSE(sse8_16bit, Src1, Src4, 16);
+                 printf("%s -   sse8_16bit#3 %.3f usec       sse=%d %s\n",
+                            cpu->name, t, sse, (sse!=146340935)?"| ERROR": "");
+                 TEST_SSE(sse8_16bit, Src2, Src3, 16);
+                 printf("%s -   sse8_16bit#4 %.3f usec       sse=%d %s\n",
+                            cpu->name, t, sse, (sse!=130136661)?"| ERROR": "");
+                 TEST_SSE(sse8_16bit, Src2, Src4, 16);
+                 printf("%s -   sse8_16bit#5 %.3f usec       sse=%d %s\n",
+                            cpu->name, t, sse, (sse!=136870353)?"| ERROR": "");
+                 TEST_SSE(sse8_16bit, Src3, Src4, 16);
+                 printf("%s -   sse8_16bit#6 %.3f usec       sse=%d %s\n",
+                            cpu->name, t, sse, (sse!=164107772)?"| ERROR": "");
+                 /* 8 bit element blocks */
+                 TEST_SSE(sse8_8bit, (int8_t*)Src1, (int8_t*)Src2, 8);
+                 printf("%s -    sse8_8bit#1 %.3f usec       sse=%d %s\n",
+                            cpu->name, t, sse, (sse!=1356423)?"| ERROR": "");
+                 TEST_SSE(sse8_8bit, (int8_t*)Src1, (int8_t*)Src3, 8);
+                 printf("%s -    sse8_8bit#2 %.3f usec       sse=%d %s\n",
+                            cpu->name, t, sse, (sse!=1173074)?"| ERROR": "");
+                 TEST_SSE(sse8_8bit, (int8_t*)Src1, (int8_t*)Src4, 8);
+                 printf("%s -    sse8_8bit#3 %.3f usec       sse=%d %s\n",
+                            cpu->name, t, sse, (sse!=1092357)?"| ERROR": "");
+                 TEST_SSE(sse8_8bit, (int8_t*)Src2, (int8_t*)Src3, 8);
+                 printf("%s -    sse8_8bit#4 %.3f usec       sse=%d %s\n",
+                            cpu->name, t, sse, (sse!=1360239)?"| ERROR": "");
+                 TEST_SSE(sse8_8bit, (int8_t*)Src2, (int8_t*)Src4, 8);
+                 printf("%s -    sse8_8bit#5 %.3f usec       sse=%d %s\n",
+                            cpu->name, t, sse, (sse!=1208414)?"| ERROR": "");
+                 TEST_SSE(sse8_8bit, (int8_t*)Src3, (int8_t*)Src4, 8);
+                 printf("%s -    sse8_8bit#6 %.3f usec       sse=%d %s\n",
+                            cpu->name, t, sse, (sse!=1099285)?"| ERROR": "");
      printf( " --- \n" );
    }
  }
-Line 580
+Line 964
   * test non-zero AC counting
   *********************************************************************/
- #define TEST_CBP(FUNC, SRC)                   \
+ #define TEST_CBP(FUNC, SRC, NB)           \
      t = gettime_usec();                       \
      emms();                                   \
-     for(tst=0; tst<nb_tests; ++tst) {         \
+ for(tst=0; tst<NB; ++tst) {         \
        cbp = (FUNC)((SRC));                    \
      }                                         \
      emms();                                   \
-Line 592
+Line 976
  void test_cbp()
  {
    const int nb_tests = 10000*speed_ref;
-   int i;
+         int i, n, m;
    CPU *cpu;
-   int16_t  Src1[6*64], Src2[6*64], Src3[6*64], Src4[6*64];
+         DECLARE_ALIGNED_MATRIX(Src1, 6, 64, int16_t, 16);
+         DECLARE_ALIGNED_MATRIX(Src2, 6, 64, int16_t, 16);
+         DECLARE_ALIGNED_MATRIX(Src3, 6, 64, int16_t, 16);
+         DECLARE_ALIGNED_MATRIX(Src4, 6, 64, int16_t, 16);
+   DECLARE_ALIGNED_MATRIX(Src5, 6, 64, int16_t, 16);
    printf( "\n =====  test cbp =====\n" );
    for(i=0; i<6*64; ++i) {
-     Src1[i] = (i*i*3/8192)&(i/64)&1;  // 'random'
+                 Src1[i] = (i*i*3/8192)&(i/64)&1;  /* 'random' */
-     Src2[i] = (i<3*64);               // half-full
+                 Src2[i] = (i<3*64);               /* half-full */
      Src3[i] = ((i+32)>3*64);
      Src4[i] = (i==(3*64+2) || i==(5*64+9));
+     Src5[i] = ieee_rand(0,1) ? -1 : 1;  /* +/- test */
    }
-   for(cpu = cpu_short_list2; cpu->name!=0; ++cpu)
+         for(cpu = cpu_list; cpu->name!=0; ++cpu)
    {
      double t;
      int tst, cbp;
-Line 613
+Line 1002
      if (!init_cpu(cpu))
        continue;
-     TEST_CBP(calc_cbp, Src1);
+                 TEST_CBP(calc_cbp, Src1, nb_tests);
-     printf( "%s -   calc_cbp#1 %.3f usec       cbp=0x%x\n", cpu->name, t, cbp );
+                 printf("%s -   calc_cbp#1 %.3f usec       cbp=0x%02x %s\n",
-     if (cbp!=0x15) printf( "*** CRC ERROR! ***\n" );
+                            cpu->name, t, cbp, (cbp!=0x15)?"| ERROR": "");
-     TEST_CBP(calc_cbp, Src2);
+                 TEST_CBP(calc_cbp, Src2, nb_tests);
-     printf( "%s -   calc_cbp#2 %.3f usec       cbp=0x%x\n", cpu->name, t, cbp );
+                 printf("%s -   calc_cbp#2 %.3f usec       cbp=0x%02x %s\n",
-     if (cbp!=0x38) printf( "*** CRC ERROR! ***\n" );
+                            cpu->name, t, cbp, (cbp!=0x38)?"| ERROR": "");
-     TEST_CBP(calc_cbp, Src3);
+                 TEST_CBP(calc_cbp, Src3, nb_tests);
-     printf( "%s -   calc_cbp#3 %.3f usec       cbp=0x%x\n", cpu->name, t, cbp );
+                 printf("%s -   calc_cbp#3 %.3f usec       cbp=0x%02x %s\n",
-     if (cbp!=0x0f) printf( "*** CRC ERROR! ***\n" );
+                            cpu->name, t, cbp, (cbp!=0x0f)?"| ERROR": "" );
-     TEST_CBP(calc_cbp, Src4);
+                 TEST_CBP(calc_cbp, Src4, nb_tests);
-     printf( "%s -   calc_cbp#4 %.3f usec       cbp=0x%x\n", cpu->name, t, cbp );
+                 printf("%s -   calc_cbp#4 %.3f usec       cbp=0x%02x %s\n",
-     if (cbp!=0x05) printf( "*** CRC ERROR! ***\n" );
+                            cpu->name, t, cbp, (cbp!=0x05)?"| ERROR": "" );
+                 TEST_CBP(calc_cbp, Src5, nb_tests);
+                 printf("%s -   calc_cbp#4 %.3f usec       cbp=0x%02x %s\n",
+                            cpu->name, t, cbp, (cbp!=0x3f)?"| ERROR": "" );
      printf( " --- \n" );
    }
+         for(cpu = cpu_list; cpu->name!=0; ++cpu)  /* bench suggested by Carlo (carlo dot bramix at libero dot it) */
+         {
+                 double t;
+                 int tst, cbp, err;
+                 if (!init_cpu(cpu))
+                         continue;
+     err = 0;
+     for(n=0; n<6; ++n)
+     {
+       for(m=0; m<64; ++m)
+       {
+         for(i=0; i<6*64; ++i)
+           Src1[i] = (i== (m + n*64));
+         TEST_CBP(calc_cbp, Src1, 1);
+         if (cbp!= (((m!=0)<<(5-n))))
+         {
+           printf( "%s -   calc_cbp#5: ERROR at pos %d / %d!\n", cpu->name, n, m);
+           err = 1;
+           break;
+         }
+       }
+     }
+     if (!err)
+       printf( " %s -    calc_cbp#5 : OK\n", cpu->name );
+         }
  }
  /*********************************************************************
-Line 755
+Line 1177
    }
  }
- //////////////////////////////////////////////////////////
+ ///* ////////////////////////////////////////////////////// */
  /* Pseudo-random generator specified by IEEE 1180 */
  static long ieee_seed = 1;
-Line 842
+Line 1264
    static const double ILimits[5] = { 1., 0.06, 0.02, 0.015, 0.0015 };
    int Loops = 10000;
    int i, m, n;
-   short Blk0[64];     // reference
+         DECLARE_ALIGNED_MATRIX(Blk0, 8, 8, short, 16); /* reference */
-   short Blk[64], iBlk[64];
+         DECLARE_ALIGNED_MATRIX(Blk,  8, 8, short, 16);
-   short Ref_FDCT[64];
+         DECLARE_ALIGNED_MATRIX(iBlk, 8, 8, short, 16);
-   short Ref_IDCT[64];
+         DECLARE_ALIGNED_MATRIX(Ref_FDCT, 8, 8, short, 16);
+         DECLARE_ALIGNED_MATRIX(Ref_IDCT, 8, 8, short, 16);
-   STATS_8x8 FStats; // forward dct stats
+         STATS_8x8 FStats; /* forward dct stats */
-   STATS_8x8 IStats; // inverse dct stats
+         STATS_8x8 IStats; /* inverse dct stats */
    CPU *cpu;
-Line 871
+Line 1294
        for(i=0; i<64; ++i)
          Blk0[i] = (short)ieee_rand(Min,Max) * Sign;
-         // hmm, I'm not quite sure this is exactly
+                         /* hmm, I'm not quite sure this is exactly */
-         // the tests described in the norm. check...
+                         /* the tests described in the norm. check... */
        memcpy(Ref_FDCT, Blk0, 64*sizeof(short));
        ref_fdct(Ref_FDCT);
-Line 899
+Line 1322
      printf( "\n  -- FDCT report --\n" );
  //    print_stats(&FStats);
-     report_stats(&FStats, 0); // so far I know, IEEE1180 says nothing for fdct
+                 report_stats(&FStats, 0); /* so far I know, IEEE1180 says nothing for fdct */
      for(i=0; i<64; i++) Blk[i] = 0;
      emms(); fdct(Blk); emms();
-Line 921
+Line 1344
  void test_dct_saturation(int Min, int Max)
  {
-     // test behaviour on input range fringe
+ /* test behaviour on input range fringe */
    int i, n, p;
    CPU *cpu;
- //  const short IDCT_MAX =  2047;  // 12bits input
+ //  const short IDCT_MAX =  2047;  /* 12bits input */
  //  const short IDCT_MIN = -2048;
- //  const short IDCT_OUT =   256;  // 9bits ouput
+ //  const short IDCT_OUT =   256;  /* 9bits ouput */
    const int Partitions = 4;
    const int Loops = 10000 / Partitions;
-Line 944
+Line 1367
      printf( "\n===== IEEE test for %s Min=%d Max=%d =====\n",
        cpu->name, Min, Max );
-               // FDCT tests //
+                 /* FDCT tests // */
      init_stats(&Stats);
-       // test each computation channels separately
+                 /* test each computation channels separately */
      for(i=0; i<64; i++) Blk[i] = Blk0[i] = ((i/8)==(i%8)) ? Max : 0;
      ref_fdct(Blk0);
      emms(); fdct(Blk); emms();
-Line 959
+Line 1382
      emms(); fdct(Blk); emms();
      store_stats(&Stats, Blk, Blk0);
-       // randomly saturated inputs
+                 /* randomly saturated inputs */
      for(p=0; p<Partitions; ++p)
      {
        for(n=0; n<Loops; ++n)
-Line 975
+Line 1398
      report_stats(&Stats, 0);
-               // IDCT tests //
+                 /* IDCT tests // */
  #if 0
-       // no finished yet
+                 /* no finished yet */
      init_stats(&Stats);
-     // test each computation channel separately
+ /* test each computation channel separately */
      for(i=0; i<64; i++) Blk[i] = Blk0[i] = ((i/8)==(i%8)) ? IDCT_MAX : 0;
      ref_idct(Blk0);
      emms(); idct(Blk); emms();
-Line 994
+Line 1417
      for(i=0; i<64; i++) { CLAMP(Blk0[i], IDCT_OUT); CLAMP(Blk[i], IDCT_OUT); }
      store_stats(&Stats, Blk, Blk0);
-       // randomly saturated inputs
+                 /* randomly saturated inputs */
      for(p=0; p<Partitions; ++p)
      {
        for(n=0; n<Loops; ++n)
-Line 1019
+Line 1442
   * measure raw decoding speed
   *********************************************************************/
- void test_dec(const char *name, int width, int height, int with_chksum)
+ void test_dec(const char *name, int width, int height, int ref_chksum)
  {
    FILE *f = 0;
    void *dechandle = 0;
    int xerr;
-         XVID_INIT_PARAM xinit;
+         xvid_gbl_init_t xinit;
-         XVID_DEC_PARAM xparam;
+         xvid_dec_create_t xparam;
-         XVID_DEC_FRAME xframe;
+         xvid_dec_frame_t xframe;
          double t = 0.;
          int nb = 0;
    uint8_t *buf = 0;
-   uint8_t *rgb_out = 0;
+         uint8_t *yuv_out = 0;
    int buf_size, pos;
    uint32_t chksum = 0;
+         int bps = (width+31) & ~31;
-         xinit.cpu_flags = XVID_CPU_MMX | XVID_CPU_FORCE;
+         memset(&xinit, 0, sizeof(xinit));
-         xvid_init(NULL, 0, &xinit, NULL);
+         xinit.cpu_flags = cpu_mask;
-         printf( "API version: %d, core build:%d\n", xinit.api_version, xinit.core_build);
+         xinit.version = XVID_VERSION;
+         xvid_global(NULL, 0, &xinit, NULL);
+         memset(&xparam, 0, sizeof(xparam));
          xparam.width = width;
          xparam.height = height;
+         xparam.version = XVID_VERSION;
          xerr = xvid_decore(NULL, XVID_DEC_CREATE, &xparam, NULL);
-         if (xerr!=XVID_ERR_OK) {
+         if (xerr==XVID_ERR_FAIL) {
-           printf("can't init decoder (err=%d)\n", xerr);
+                 printf("ERROR: can't init decoder (err=%d)\n", xerr);
            return;
          }
          dechandle = xparam.handle;
-Line 1051
+Line 1477
          f = fopen(name, "rb");
    if (f==0) {
-     printf( "can't open file '%s'\n", name);
+                 printf( "ERROR: can't open file '%s'\n", name);
      return;
    }
    fseek(f, 0, SEEK_END);
    buf_size = ftell(f);
    fseek(f, 0, SEEK_SET);
    if (buf_size<=0) {
-     printf("error while stating file\n");
+                 printf("ERROR: error while stating file\n");
      goto End;
    }
-   else printf( "Input size: %d\n", buf_size);
-   buf = malloc(buf_size); // should be enuf'
+         buf = malloc(buf_size);
-   rgb_out = calloc(4, width*height);  // <-room for _RGB24
+         yuv_out = calloc(1, bps*height*3/2 + 15);
-   if (buf==0 || rgb_out==0) {
+         if (buf==0 || yuv_out==0) {
-     printf( "malloc failed!\n" );
+                 printf( "ERROR: malloc failed!\n" );
      goto End;
    }
    if (fread(buf, buf_size, 1, f)!=1) {
-     printf( "file-read failed\n" );
+                 printf( "ERROR: file-read failed\n" );
      goto End;
    }
-Line 1079
+Line 1504
    pos = 0;
    t = -gettime_usec();
    while(1) {
+           int y;
+                 memset(&xframe, 0, sizeof(xframe));
+                 xframe.version = XVID_VERSION;
      xframe.bitstream = buf + pos;
      xframe.length = buf_size - pos;
-     xframe.image = rgb_out;
+                 xframe.output.plane[0] = (uint8_t*)(((size_t)yuv_out + 15) & ~15);
-     xframe.stride = width;
+                 xframe.output.plane[1] = (uint8_t*)xframe.output.plane[0] + bps*height;
-     xframe.colorspace = XVID_CSP_RGB24;
+                 xframe.output.plane[2] = (uint8_t*)xframe.output.plane[1] + bps/2;
+                 xframe.output.stride[0] = bps;
+                 xframe.output.stride[1] = bps;
+                 xframe.output.stride[2] = bps;
+                 xframe.output.csp = XVID_CSP_I420;
      xerr = xvid_decore(dechandle, XVID_DEC_DECODE, &xframe, 0);
+                 if (xerr<0) {
+                         printf("ERROR: decoding failed for frame #%d (err=%d)!\n", nb, xerr);
+                         break;
+                 }
+                 else if (xerr==0)
+                   break;
+     else if (verbose>0) printf("#%d %d\n", nb, xerr );
+                 pos += xerr;
      nb++;
-     pos += xframe.length;
-     if (with_chksum) {
+     for(y=0; y<height/2; ++y) {
-       int k = width*height;
+                   chksum = calc_crc((uint8_t*)xframe.output.plane[0] + (2*y+0)*bps, width, chksum);
-       uint32_t *ptr = (uint32_t *)rgb_out;
+                         chksum = calc_crc((uint8_t*)xframe.output.plane[0] + (2*y+1)*bps, width, chksum);
-       while(k-->0) chksum += *ptr++;
+                         chksum = calc_crc((uint8_t*)xframe.output.plane[1] + y*bps, width/2, chksum);
+                         chksum = calc_crc((uint8_t*)xframe.output.plane[2] + y*bps, width/2, chksum);
      }
      if (pos==buf_size)
        break;
-     if (xerr!=XVID_ERR_OK) {
-           printf("decoding failed for frame #%d (err=%d)!\n", nb, xerr);
-           break;
-         }
    }
    t += gettime_usec();
+         if (ref_chksum==0) {
    if (t>0.)
-     printf( "%d frames decoded in %.3f s -> %.1f FPS\n", nb, t*1.e-6f, (float)(nb*1.e6f/t) );
+                   printf( "%d frames decoded in %.3f s -> %.1f FPS   Checksum:0x%.8x\n", nb, t*1.e-6f, (float)(nb*1.e6f/t), chksum );
-   if (with_chksum)
+   }
-     printf("checksum: 0x%.8x\n", chksum);
+   else {
+                 printf("FPS:%.1f Checksum: 0x%.8x Expected:0x%.8x | %s\n",
+                   t>0. ? (float)(nb*1.e6f/t) : 0.f, chksum, ref_chksum, (chksum==ref_chksum) ? "OK" : "ERROR");
+   }
  End:
-   if (rgb_out!=0) free(rgb_out);
+         if (yuv_out!=0) free(yuv_out);
    if (buf!=0) free(buf);
    if (dechandle!=0) {
      xerr= xvid_decore(dechandle, XVID_DEC_DESTROY, NULL, NULL);
-     if (xerr!=XVID_ERR_OK)
+                 if (xerr==XVID_ERR_FAIL)
-             printf("destroy-decoder failed (err=%d)!\n", xerr);
+                         printf("ERROR: destroy-decoder failed (err=%d)!\n", xerr);
    }
    if (f!=0) fclose(f);
  }
-Line 1123
+Line 1566
  void test_bugs1()
  {
    CPU *cpu;
+         uint16_t mpeg_quant_matrices[64*8];
    printf( "\n =====  (de)quant4_intra saturation bug? =====\n" );
-   for(cpu = cpu_short_list; cpu->name!=0; ++cpu)
+         for(cpu = cpu_list; cpu->name!=0; ++cpu)
    {
      int i;
      int16_t  Src[8*8], Dst[8*8];
-Line 1135
+Line 1579
        continue;
      for(i=0; i<64; ++i) Src[i] = i-32;
-     set_intra_matrix( get_default_intra_matrix() );
+                 set_intra_matrix( mpeg_quant_matrices, get_default_intra_matrix() );
-     dequant4_intra(Dst, Src, 31, 5);
+                 dequant_mpeg_intra(Dst, Src, 31, 5, mpeg_quant_matrices);
-     printf( "dequant4_intra with CPU=%s:  ", cpu->name);
+                 printf( "dequant_mpeg_intra with CPU=%s:  ", cpu->name);
      printf( "  Out[]= " );
      for(i=0; i<64; ++i) printf( "[%d]", Dst[i]);
      printf( "\n" );
-Line 1145
+Line 1589
    printf( "\n =====  (de)quant4_inter saturation bug? =====\n" );
-   for(cpu = cpu_short_list; cpu->name!=0; ++cpu)
+         for(cpu = cpu_list; cpu->name!=0; ++cpu)
    {
      int i;
      int16_t  Src[8*8], Dst[8*8];
-Line 1154
+Line 1598
        continue;
      for(i=0; i<64; ++i) Src[i] = i-32;
-     set_inter_matrix( get_default_inter_matrix() );
+                 set_inter_matrix( mpeg_quant_matrices, get_default_inter_matrix() );
-     dequant4_inter(Dst, Src, 31);
+                 dequant_mpeg_inter(Dst, Src, 31, mpeg_quant_matrices);
-     printf( "dequant4_inter with CPU=%s:  ", cpu->name);
+                 printf( "dequant_mpeg_inter with CPU=%s:  ", cpu->name);
      printf( "  Out[]= " );
      for(i=0; i<64; ++i) printf( "[%d]", Dst[i]);
      printf( "\n" );
-Line 1166
+Line 1610
  void test_dct_precision_diffs()
  {
    CPU *cpu;
-   short Blk[8*8], Blk0[8*8];
+         DECLARE_ALIGNED_MATRIX(Blk, 8, 8, int16_t, 16);
+         DECLARE_ALIGNED_MATRIX(Blk0, 8, 8, int16_t, 16);
    printf( "\n =====  fdct/idct precision diffs =====\n" );
-   for(cpu = cpu_short_list; cpu->name!=0; ++cpu)
+         for(cpu = cpu_list; cpu->name!=0; ++cpu)
    {
      int i;
-Line 1199
+Line 1644
    const int max_Q = 31;
    int i, n, qm, q;
    CPU *cpu;
-   int16_t  Src[8*8], Dst[8*8];
+         DECLARE_ALIGNED_MATRIX(Src, 8, 8, int16_t, 16);
+         DECLARE_ALIGNED_MATRIX(Dst, 8, 8, int16_t, 16);
    uint8_t Quant[8*8];
    CPU cpu_bug_list[] = { { "PLAINC", 0 }, { "MMX   ", XVID_CPU_MMX }, {0,0} };
    uint16_t Crcs_Inter[2][32];
    uint16_t Crcs_Intra[2][32];
+         DECLARE_ALIGNED_MATRIX(mpeg_quant_matrices, 8, 64, uint16_t, 16);
    printf( "\n =====  test MPEG4-quantize bug =====\n" );
    for(i=0; i<64; ++i) Src[i] = 2048*(i-32)/32;
-Line 1212
+Line 1660
    for(qm=1; qm<=255; ++qm)
    {
      for(i=0; i<8*8; ++i) Quant[i] = qm;
-     set_inter_matrix( Quant );
+                 set_inter_matrix( mpeg_quant_matrices, Quant );
      for(n=0, cpu = cpu_bug_list; cpu->name!=0; ++cpu, ++n)
      {
-Line 1223
+Line 1671
        for(q=1; q<=max_Q; ++q) {
          emms();
-         quant4_inter( Dst, Src, q );
+                                 quant_mpeg_inter( Dst, Src, q, mpeg_quant_matrices );
          emms();
          for(s=0, i=0; i<64; ++i) s+=((uint16_t)Dst[i])^i;
          Crcs_Inter[n][q] = s;
-Line 1242
+Line 1690
    for(qm=1; qm<=255; ++qm)
    {
      for(i=0; i<8*8; ++i) Quant[i] = qm;
-     set_intra_matrix( Quant );
+                 set_intra_matrix( mpeg_quant_matrices, Quant );
      for(n=0, cpu = cpu_bug_list; cpu->name!=0; ++cpu, ++n)
      {
-Line 1253
+Line 1701
        for(q=1; q<=max_Q; ++q) {
          emms();
-         quant4_intra( Dst, Src, q, q);
+                                 quant_mpeg_intra( Dst, Src, q, q, mpeg_quant_matrices);
          emms();
          for(s=0, i=0; i<64; ++i) s+=((uint16_t)Dst[i])^i;
          Crcs_Intra[n][q] = s;
-Line 1270
+Line 1718
  }
  /*********************************************************************
+  * test some YUV func
+  *********************************************************************/
+ #define ENTER \
+ for(i=0; i<(int)sizeof(Dst0); ++i) Dst0[0][i] = 0;   \
+ t = gettime_usec();                   \
+ emms();
+ #define LEAVE \
+ emms();                             \
+ t = (gettime_usec() - t) / nb_tests;  \
+         iCrc = calc_crc((uint8_t*)Dst0, sizeof(Dst0), CRC32_INITIAL)
+ #define TEST_YUYV(FUNC, S, FLIP)                \
+ ENTER                               \
+ for(tst=0; tst<nb_tests; ++tst) (FUNC)(Dst0[0], S*WIDTH, Src0[0], Src0[1], Src0[2], WIDTH, WIDTH/2, WIDTH, HEIGHT, (FLIP)); \
+ LEAVE
+ static const int yuv_CRCs[6][2] = {
+         {0x0f4fb96b,0x780b6a68}
+ ,       {0xa986b289,0x65e49b76}
+ ,       {0x7f19c152,0xd539b86e}
+ ,       {0x0f4fb96b,0x780b6a68}
+ ,       {0xa986b289,0x65e49b76}
+ ,       {0x36ab8b57,0x1cd92fee}
+ };
+ #define WIDTH 128
+ #define HEIGHT 32
+ void test_yuv()
+ {
+         const int nb_tests = 200*speed_ref;
+         CPU *cpu;
+         uint8_t Src0[3][WIDTH*HEIGHT];
+         uint8_t Dst0[4][WIDTH*HEIGHT];
+         int i, j, with_flip;
+         double t;
+         int tst, iCrc;
+         colorspace_init();
+         ieee_reseed(1);
+         for(i=0; i<(int)sizeof(Src0); ++i) Src0[0][i] = ieee_rand(0,255);
+         for(i=0; i<(int)sizeof(Dst0); ++i) Dst0[0][i] = 0x5a;
+         printf( "\n ===  test YUV ===\n" );
+         for(with_flip=0; with_flip<=1; ++with_flip) {
+                 init_cpu(&cpu_list[0]);
+                 TEST_YUYV(yv12_to_yuyv_c, 4, with_flip);
+                 printf(" yv12_to_yuyv_c %.3f usec       crc32=0x%08x %s\n",
+                            t, iCrc, (iCrc!=yuv_CRCs[0][with_flip])?"| ERROR": "" );
+                 TEST_YUYV(yv12_to_uyvy_c, 4, with_flip);
+                 printf(" yv12_to_uyvy_c %.3f usec       crc32=0x%08x %s\n",
+                         t, iCrc, (iCrc!=yuv_CRCs[1][with_flip])?"| ERROR": "" );
+                 TEST_YUYV(yv12_to_bgra_c, 4, with_flip);
+                 printf(" yv12_to_bgra_c %.3f usec       crc32=0x%08x %s\n",
+                         t, iCrc, (iCrc!=yuv_CRCs[2][with_flip])?"| ERROR": "" );
+ #if defined(ARCH_IS_IA32) || defined(ARCH_IS_X86_64)
+                 init_cpu(&cpu_list[1]);
+                 TEST_YUYV(yv12_to_yuyv_mmx, 4, with_flip);
+                 printf(" yv12_to_yuyv_mmx %.3f usec       crc32=0x%08x %s\n",
+                         t, iCrc, (iCrc!=yuv_CRCs[3][with_flip])?"| ERROR": "" );
+                 TEST_YUYV(yv12_to_uyvy_mmx, 4, with_flip);
+                 printf(" yv12_to_uyvy_mmx %.3f usec       crc32=0x%08x %s\n",
+                         t, iCrc, (iCrc!=yuv_CRCs[4][with_flip])?"| ERROR": "" );
+                 TEST_YUYV(yv12_to_bgra_mmx, 4, with_flip);
+                 printf(" yv12_to_bgra_mmx %.3f usec       crc32=0x%08x %s\n",
+                         t, iCrc, (iCrc!=yuv_CRCs[5][with_flip])?"| ERROR": "" );
+ #endif
+ #ifdef ARCH_IS_PPC
+                 init_cpu(&cpu_list[1]);
+                 TEST_YUYV(yv12_to_yuyv_altivec_c, 4, with_flip);
+                 printf(" yv12_to_yuyv_altivec_c %.3f usec       crc32=0x%08x %s\n",
+                         t, iCrc, (iCrc!=yuv_CRCs[3][with_flip])?"| ERROR": "" );
+                 TEST_YUYV(yv12_to_uyvy_altivec_c, 4, with_flip);
+                 printf(" yv12_to_uyvy_altivec_c %.3f usec       crc32=0x%08x %s\n",
+                         t, iCrc, (iCrc!=yuv_CRCs[4][with_flip])?"| ERROR": "" );
+ #endif
+         }
+         printf( " --- \n" );
+ }
+ #define TEST_YV2(FUNC, WITH_UV, WITH_FLIP)        \
+ ENTER                               \
+ for(tst=0; tst<nb_tests; ++tst) (FUNC)(Dst0[0], Dst0[1], Dst0[2], WIDTH, WIDTH, \
+         Src0[0], (WITH_UV) ? Src0[1] : 0, (WITH_UV) ? Src0[2] : 0,  WIDTH, WIDTH, \
+         WIDTH-2, HEIGHT-2, WITH_FLIP); \
+ LEAVE
+ #define PRINT_NxN(DATA,W,H,STR)   {   \
+         int i,j; \
+         for(j=0; j<(H); ++j) { \
+                 for(i=0; i<(W); ++i) printf( "0x%.2x ", (DATA)[i+j*(STR)] );\
+                 printf("\n"); \
+         } \
+         printf("---\n"); \
+ }
+ static const int yv12_CRCs[2][2] = {
+         {0x5cab7cf0,0xdab46541}
+ ,       {0xe8bae865,0x1faf77b7}
+ };
+ void test_yuv2()
+ {
+         const int nb_tests = 800*speed_ref;
+         CPU *cpu;
+         uint8_t Src0[3][WIDTH*HEIGHT];
+         uint8_t Dst0[3][WIDTH*HEIGHT];
+         int with_uv, with_flip;
+         int i, j;
+         double t;
+         int tst, iCrc;
+         colorspace_init();
+         ieee_reseed(1);
+         for(i=0; i<(int)sizeof(Src0); ++i) Src0[0][i] = ieee_rand(0,255);
+         printf( "\n ===  test YV2 ===\n" );
+         for(with_flip=0; with_flip<=1; ++with_flip) {
+                 for(with_uv=0; with_uv<=1; ++with_uv) {
+                         init_cpu(&cpu_list[0]);
+                         TEST_YV2(yv12_to_yv12_c, with_uv, with_flip);
+                         printf(" yv12_to_yv12_c   %.3f usec      \tcrc32=0x%08x %s\n",
+                                 t, iCrc, (iCrc!=yv12_CRCs[with_flip][with_uv])?"| ERROR": "" );
+                         /* if (!with_uv) PRINT_NxN(Dst0[1], WIDTH/2, HEIGHT/2, WIDTH ); */
+ #if defined(ARCH_IS_IA32) || defined(ARCH_IS_X86_64)
+                         init_cpu(&cpu_list[1]);
+                         TEST_YV2(yv12_to_yv12_mmx, with_uv, with_flip);
+                         printf(" yv12_to_yv12_mmx %.3f usec     \tcrc32=0x%08x %s\n",
+                                 t, iCrc, (iCrc!=yv12_CRCs[with_flip][with_uv])?"| ERROR": "" );
+                         /* if (!with_uv) PRINT_NxN(Dst0[1], WIDTH/2, HEIGHT/2, WIDTH ); */
+                         TEST_YV2(yv12_to_yv12_xmm, with_uv, with_flip);
+                         printf(" yv12_to_yv12_xmm %.3f usec     \tcrc32=0x%08x %s\n",
+                                 t, iCrc, (iCrc!=yv12_CRCs[with_flip][with_uv])?"| ERROR": "" );
+ #endif
+                 }
+                 printf( " --- \n" );
+         }
+         printf( " ===== \n" );
+ }
+ #undef WIDTH
+ #undef HEIGHT
+ #undef ENTER
+ #undef LEAVE
+ /*********************************************************************/
+ static uint32_t __inline log2bin_v1(uint32_t value)
+ {
+   int n = 0;
+   while (value) {
+     value >>= 1;
+     n++;
+   }
+   return n;
+ }
+ static const uint8_t log2_tab_16[16] =  { 0, 1, 2, 2, 3, 3, 3, 3, 4, 4, 4, 4, 4, 4, 4, 4 };
+ static uint32_t __inline log2bin_v2(uint32_t value)
+ {
+   int n = 0;
+   if (value & 0xffff0000) {
+     value >>= 16;
+     n += 16;
+   }
+   if (value & 0xff00) {
+     value >>= 8;
+     n += 8;
+   }
+   if (value & 0xf0) {
+     value >>= 4;
+     n += 4;
+   }
+  return n + log2_tab_16[value];
+ }
+ void test_log2bin()
+ {
+         const int nb_tests = 3000*speed_ref;
+   int n, crc1=0, crc2=0;
+   uint32_t s, s0;
+   double t1, t2;
+   t1 = gettime_usec();
+   s0 = (int)(t1*31.241);
+   for(s=s0, n=0; n<nb_tests; ++n, s=(s*12363+31)&0x7fffffff)
+     crc1 += log2bin_v1(s);
+   t1 = (gettime_usec()-t1) / nb_tests;
+   t2 = gettime_usec();
+   for(s=s0, n=0; n<nb_tests; ++n, s=(s*12363+31)&0x7fffffff)
+     crc2 += log2bin_v2(s);
+   t2 = (gettime_usec() - t2) / nb_tests;
+   printf( "log2bin_v1: %.3f sec  crc=%d\n", t1, crc1 );
+   printf( "log2bin_v2: %.3f sec  crc=%d\n", t2, crc2 );
+   if (crc1!=crc2) printf( " CRC ERROR !\n" );
+ }
+ /*********************************************************************/
+ static void __inline old_gcd(int *num, int *den)
+ {
+  int i = *num;
+   while (i > 1) {
+     if (*num % i == 0 && *den % i == 0) {
+       *num /= i;
+       *den /= i;
+       i = *num;
+       continue;
+     }
+     i--;
+   }
+ }
+ static uint32_t gcd(int num, int den)
+ {
+   int tmp;
+   while( (tmp=num%den) ) { num = den; den = tmp; }
+   return den;
+ }
+ static void __inline new_gcd(int *num, int *den)
+ {
+   const int div = gcd(*num, *den);
+   if (num) {
+     *num /= div;
+     *den /= div;
+   }
+ }
+ void test_gcd()
+ {
+   const int nb_tests = 10*speed_ref;
+   int i;
+   uint32_t crc1=0, crc2=0;
+   uint32_t n0, n, d0, d;
+   double t1, t2;
+   t1 = gettime_usec();
+   n0 = 0xfffff & (int)(t1*31.241);
+   d0 = 0xfffff & (int)( ((n0*4123)%17) | 1 );
+   for(n=n0, d=d0, i=0; i<nb_tests; ++i) {
+     old_gcd(&n, &d);
+     crc1 = (((crc1>>4)^d) + ((crc1<<2)^n) ) & 0xffffff;
+     n = d;
+     d = (d*12363+31) & 0xffff;
+     d |= !d;
+   }
+   t1 = (gettime_usec()-t1) / nb_tests;
+   t2 = gettime_usec();
+   for(n=n0, d=d0, i=0; i<nb_tests; ++i) {
+     new_gcd(&n, &d);
+     crc2 = (((crc2>>4)^d) + ((crc2<<2)^n) ) & 0xffffff;
+     n = d;
+     d = (d*12363+31) & 0xffff;
+     d |= !d;
+   }
+   t2 = (gettime_usec() - t2) / nb_tests;
+   printf( "old_gcd: %.3f sec  crc=%d\n", t1, crc1 );
+   printf( "new_gcd: %.3f sec  crc=%d\n", t2, crc2 );
+   if (crc1!=crc2) printf( " CRC ERROR !\n" );
+ }
+ /*********************************************************************
+  * test compiler
+  *********************************************************************/
+ void test_compiler() {
+   int nb_err = 0;
+   int32_t v;
+   if (sizeof(uint16_t)<2) {
+     printf( "ERROR: sizeof(uint16_t)<2 !!\n" );
+     nb_err++;
+   }
+   if (sizeof(int16_t)<2) {
+     printf( "ERROR: sizeof(int16_t)<2 !!\n" );
+     nb_err++;
+   }
+   if (sizeof(uint8_t)!=1) {
+     printf( "ERROR: sizeof(uint8_t)!=1 !!\n" );
+     nb_err++;
+   }
+   if (sizeof(int8_t)!=1) {
+     printf( "ERROR: sizeof(int8_t)!=1 !!\n" );
+     nb_err++;
+   }
+   if (sizeof(uint32_t)<4) {
+     printf( "ERROR: sizeof(uint32_t)<4 !!\n" );
+     nb_err++;
+   }
+   if (sizeof(int32_t)<4) {
+     printf( "ERROR: sizeof(int32_t)<4 !!\n" );
+     nb_err++;
+   }
+          /* yes, i know, this test is silly. But better be safe than sorry. :) */
+   for(v=1000; v>=0; v--) {
+     if ( (v>>2) != v/4)
+       nb_err++;
+   }
+   for(v=-1000; v!=-1; v++) {
+     if ( (v>>2) != (v/4)-!!(v%4))
+       nb_err++;
+   }
+   if (nb_err!=0) {
+     printf( "ERROR! please post your platform/compiler specs to xvid-devel@xvid.org !\n" );
+   }
+ }
+ /*********************************************************************
+  * test SSIM functions
+  *********************************************************************/
+ typedef int (*lumfunc)(uint8_t* ptr, int stride);
+ typedef void (*csfunc)(uint8_t* ptro, uint8_t* ptrc, int stride, int lumo, int lumc, int* pdevo, int* pdevc, int* pcorr);
+ extern int lum_8x8_c(uint8_t* ptr, int stride);
+ extern int lum_8x8_mmx(uint8_t* ptr, int stride);
+ extern int lum_2x8_c(uint8_t* ptr, int stride);
+ extern void consim_c(uint8_t* ptro, uint8_t* ptrc, int stride, int lumo, int lumc, int* pdevo, int* pdevc, int* pcorr);
+ extern void consim_mmx(uint8_t* ptro, uint8_t* ptrc, int stride, int lumo, int lumc, int* pdevo, int* pdevc, int* pcorr);
+ extern void consim_sse2(uint8_t* ptro, uint8_t* ptrc, int stride, int lumo, int lumc, int* pdevo, int* pdevc, int* pcorr);
+ void test_SSIM()
+ {
+         const int nb_tests = 3000*speed_ref;
+         int tst;
+         CPU *cpu;
+         int i;
+         int devs[3];
+         long lumo, lumc;
+         DECLARE_ALIGNED_MATRIX(Ref1, 16, 16, uint8_t, 16);
+         DECLARE_ALIGNED_MATRIX(Ref2, 16, 16, uint8_t, 16);
+         lumfunc lum8x8;
+         lumfunc lum2x8;
+         csfunc  csim;
+         ieee_reseed(1);
+         printf( "\n ======  test SSIM ======\n" );
+         for(i=0; i<16*16;++i) {
+                 long v1, v2;
+                 v1 = ieee_rand(-256, 511);
+                 v2 = ieee_rand(-256, 511);
+                 Ref1[i] = (v1<0) ? 0 : (v1>255) ? 255 : v1;
+                 Ref2[i] = (v2<0) ? 0 : (v2>255) ? 255 : v2;
+         }
+         lumc = ieee_rand(0, 255);
+         lumo = ieee_rand(0, 255);
+         for(cpu = cpu_list; cpu->name!=0; ++cpu)
+         {
+                 double t;
+                 int m;
+                 if (!init_cpu(cpu))
+                         continue;
+                 lum8x8 = lum_8x8_c;
+                 lum2x8 = lum_2x8_c;
+                 csim   = consim_c;
+ #if defined(ARCH_IS_IA32) || defined(ARCH_IS_X86_64)
+                 if (cpu->cpu & XVID_CPU_MMX){
+                         lum8x8 = lum_8x8_mmx;
+                         csim = consim_mmx;
+                 }
+                 if (cpu->cpu & XVID_CPU_MMX){
+                         csim = consim_sse2;
+                 }
+ #endif
+                 t = gettime_usec();
+                 emms();
+                 for(tst=0; tst<nb_tests; ++tst) m = lum8x8(Ref1, 16);
+                 emms();
+                 t = (gettime_usec() - t) / nb_tests;
+                 printf("%s - ssim-lum8x8    %.3f usec       m=%d %s\n",
+                            cpu->name, t, m,
+                            (m!=8230)?"| ERROR": "" );
+                 t = gettime_usec();
+                 emms();
+                 for(tst=0; tst<nb_tests; ++tst) m = lum2x8(Ref1+8, 16);
+                 emms();
+                 t = (gettime_usec() - t) / nb_tests;
+                 printf("%s - ssim-lum2x8    %.3f usec       m=%d %s\n",
+                            cpu->name, t, m,
+                            (m!=681)?"| ERROR": "" );
+                 t = gettime_usec();
+                 emms();
+                 for(tst=0; tst<nb_tests; ++tst) csim(Ref1, Ref2, 16, lumo, lumc, devs+0, devs+1, devs+2);
+                 emms();
+                 t = (gettime_usec() - t) / nb_tests;
+                 printf("%s - ssim-consim    %.3f usec       devs=[0x%x 0x%x 0x%x] %s\n",
+                            cpu->name, t, devs[0], devs[1], devs[2],
+                            (devs[0]!=0x1bdf0f || devs[1]!=0x137258 ||  devs[2]!=0xcdb13)?"| ERROR": "" );
+                 printf( " --- \n" );
+         }
+ }
+ /*********************************************************************
+  * test bitstream functions
+  *********************************************************************/
+ #define BIT_BUF_SIZE 2000
+ static void test_bits()
+ {
+   const int nb_tests = 50*speed_ref;
+   int tst;
+   uint32_t Crc;
+   uint8_t Buf[BIT_BUF_SIZE];
+   uint32_t Extracted[BIT_BUF_SIZE*8]; /* worst case: bits read 1 by 1 */
+   int Lens[BIT_BUF_SIZE*8];
+   double t1;
+   printf( "\n ===  test bitstream ===\n" );
+   ieee_reseed(1);
+   Crc = 0;
+   t1 = gettime_usec();
+   for(tst=0; tst<nb_tests; ++tst) {
+         Bitstream bs;
+         int m, m2, l, l2;
+         for(l=0; l<BIT_BUF_SIZE; ++l)
+                 Buf[l] = (uint8_t)ieee_rand(0,255);
+         l = BIT_BUF_SIZE - ieee_rand(1,BIT_BUF_SIZE/10);
+         BitstreamInit(&bs, (void*)(Buf+BIT_BUF_SIZE-l), l);
+         BitstreamReset(&bs);
+         for(l2=l*8, m=0; l2>0; m++) {
+                 const int b = ieee_rand(1,32);
+                 Lens[m] = b;
+                 l2 -= b;
+                 if (l2<0) break;
+                 Extracted[m] = BitstreamShowBits(&bs, b);
+                 BitstreamSkip(&bs, b);
+ //              printf( "<= %d: %d 0x%x\n", m, b, Extracted[m]);
+         }
+         BitstreamReset(&bs);
+         for(m2=0; m2<m; ++m2) {
+                 const int b = Lens[m2];
+                 const uint32_t v = BitstreamGetBits(&bs, b);
+                 Crc |= (v!=Extracted[m2]);
+ //              printf( "=> %d: %d 0x%x %c\n", m2, b, v, " *"[Crc]);
+         }
+   }
+   t1 = (gettime_usec() - t1) / nb_tests;
+   printf(" test_bits   %.3f usec   %s\n", t1, (Crc!=0)?"| ERROR": "" );
+ }
+ /*********************************************************************
   * main
   *********************************************************************/
- int main(int argc, char *argv[])
+ static void arg_missing(const char *opt)
+ {
+   printf( "missing argument after option '%s'\n", opt);
+   exit(-1);
+ }
+ int main(int argc, const char *argv[])
+ {
+         int c, what = 0;
+         int width, height;
+         uint32_t chksum = 0;
+         const char * test_bitstream = 0;
+ #if defined(WIN32) && defined(ARCH_IS_X86_64)
+         DECLARE_ALIGNED_MATRIX(xmm_save, 2, 4, uint64_t, 16);
+         // assumes xmm6 and xmm7 won't be falsely preserved by C code
+         for(c=0;c<4;c++)
+                 xmm_save[c] = read_counter();
+         prime_xmm(xmm_save);
+ #endif
+         cpu_mask = 0;  // default => will use autodectect
+         for(c=1; c<argc; ++c)
  {
-   int what = 0;
+           if (!strcmp(argv[c], "-v")) verbose++;
-   if (argc>1) what = atoi(argv[1]);
+           else if (!strcmp(argv[c], "-c"))      cpu_mask = 0 /* PLAIN_C */ | XVID_CPU_FORCE;
+           else if (!strcmp(argv[c], "-mmx"))    cpu_mask = XVID_CPU_MMX    | XVID_CPU_FORCE;
+           else if (!strcmp(argv[c], "-mmxext")) cpu_mask = XVID_CPU_MMXEXT | XVID_CPU_MMX | XVID_CPU_FORCE;
+           else if (!strcmp(argv[c], "-sse2"))   cpu_mask = XVID_CPU_SSE2   | XVID_CPU_MMXEXT | XVID_CPU_MMX | XVID_CPU_FORCE;
+           else if (!strcmp(argv[c], "-sse3"))   cpu_mask = XVID_CPU_SSE3   | XVID_CPU_SSE2 | XVID_CPU_MMXEXT | XVID_CPU_MMX | XVID_CPU_FORCE;
+           else if (!strcmp(argv[c], "-sse4"))   cpu_mask = XVID_CPU_SSE41  | XVID_CPU_SSE3 | XVID_CPU_SSE2 | XVID_CPU_MMXEXT | XVID_CPU_MMX | XVID_CPU_FORCE;
+       else if (!strcmp(argv[c], "-3dnow"))  cpu_mask = XVID_CPU_3DNOW  | XVID_CPU_FORCE;
+           else if (!strcmp(argv[c], "-3dnowe")) cpu_mask = XVID_CPU_3DNOW  | XVID_CPU_3DNOWEXT | XVID_CPU_FORCE;
+           else if (!strcmp(argv[c], "-altivec")) cpu_mask = XVID_CPU_ALTIVEC | XVID_CPU_FORCE;
+           else if (!strcmp(argv[c], "-spd")) {
+       if (++c==argc) arg_missing( argv[argc-1] );
+       speed_ref = atoi(argv[c]);
+     }
+           else if (argv[c][0]!='-') {
+             what = atoi(argv[c]);
+             if (what==9) {
+               if (c+4>argc) {
+                 printf("usage: %s %d bitstream width height (checksum)\n", argv[0], what);
+                 exit(-1);
+         }
+         test_bitstream = argv[++c];
+               width  = atoi(argv[++c]);
+               height = atoi(argv[++c]);
+               if (c+1<argc && argv[c+1][0]!='-') {
+                 if (sscanf(argv[c+1], "0x%x", &chksum)!=1) {
+                   printf( "can't read checksum value.\n" );
+                   exit(-1);
+           }
+           else c++;
+         }
+ //        printf( "[%s] %dx%d (0x%.8x)\n", test_bitstream, width, height, chksum);
+       }
+     }
+     else {
+       printf( "unrecognized option '%s'\n", argv[c]);
+       exit(-1);
+     }
+   }
    if (what==0 || what==1) test_dct();
    if (what==0 || what==2) test_mb();
    if (what==0 || what==3) test_sad();
    if (what==0 || what==4) test_transfer();
    if (what==0 || what==5) test_quant();
    if (what==0 || what==6) test_cbp();
+         if (what==0 || what==10) test_sse();
+         if (what==0 || what==11) test_log2bin();
+         if (what==0 || what==12) test_gcd();
+         if (what==0 || what==13) test_compiler();
+         if (what==0 || what==14) test_yuv();
+         if (what==0 || what==15) test_SSIM();
+         if (what==0 || what==16) test_yuv2();
+         if (what==0 || what==17) test_bits();
    if (what==7) {
      test_IEEE1180_compliance(-256, 255, 1);
- #if 0
      test_IEEE1180_compliance(-256, 255,-1);
      test_IEEE1180_compliance(  -5,   5, 1);
      test_IEEE1180_compliance(  -5,   5,-1);
      test_IEEE1180_compliance(-300, 300, 1);
      test_IEEE1180_compliance(-300, 300,-1);
- #endif
    }
    if (what==8) test_dct_saturation(-256, 255);
-   if (what==9) {
+         if (test_bitstream)
-     int width, height;
+           test_dec(test_bitstream, width, height, chksum);
-     if (argc<5) {
-       printf("usage: %s %d [bitstream] [width] [height]\n", argv[0], what);
-       return 1;
-     }
-     width = atoi(argv[3]);
-     height = atoi(argv[4]);
-     test_dec(argv[2], width, height, (argc>5));
-   }
    if (what==-1) {
      test_dct_precision_diffs();
      test_bugs1();
-Line 1314
+Line 2288
    if (what==-2)
      test_quant_bug();
-   return 0;
+ #if defined(WIN32) && defined(ARCH_IS_X86_64)
+         get_xmm(xmm_save+4);
+         if (memcmp(xmm_save, xmm_save+4, 4*sizeof(int64_t))) {
+                 printf("\nWIN64 ERROR: XMM6 and XMM7 contents not preserved!\n"
+                        "        XMM6                             XMM7\n"
+                        "Before: %.16I64X%.16I64X %.16I64X%.16I64X\n"
+                        "After:  %.16I64X%.16I64X %.16I64X%.16I64X",
+                         xmm_save[0],xmm_save[1],xmm_save[2],xmm_save[3],
+                         xmm_save[4],xmm_save[5],xmm_save[6],xmm_save[7]);
+         } else {
+                 printf("\nWIN64: XMM6 and XMM7 contents preserved correctly.\n");
  }
+ #endif
- /*********************************************************************
+         if ((what >= 0 && what <= 6) || what == 10) {
-  * 'Reference' output (except for timing) on a PIII 1.13Ghz/linux
+                 printf("\n\n"
-  *********************************************************************/
+                            "NB: If a function isn't optimised for a specific set of intructions,\n"
+                            "    a C function is used instead. So don't panic if some functions\n"
+                            "    may appear to be slow.\n");
+         }
-     /* as of 07/01/2002, there's a problem with mpeg4-quantization */
+ #if defined(ARCH_IS_IA32) || defined(ARCH_IS_X86_64)
- /*
+         if (what == 0 || what == 5) {
+                 printf("\n"
+                            "NB: MMX mpeg4 quantization is known to have very small errors (+/-1 magnitude)\n"
+                            "    for 1 or 2 coefficients a block. This is mainly caused by the fact the unit\n"
+                            "    test goes far behind the usual limits of real encoding. Please do not report\n"
+                            "    this error to the developers.\n");
+         }
+ #endif
-  ===== test fdct/idct =====
+         return 0;
- PLAINC -  3.312 usec       PSNR=13.291  MSE=3.000
+ }
- MMX    -  0.591 usec       PSNR=13.291  MSE=3.000
- MMXEXT -  0.577 usec       PSNR=13.291  MSE=3.000
- SSE2   -  0.588 usec       PSNR=13.291  MSE=3.000
-DNOW  - skipped...
-DNOWE - skipped...
-  ===  test block motion ===
- PLAINC - interp- h-round0 0.911 usec       iCrc=8107
- PLAINC -           round1 0.863 usec       iCrc=8100
- PLAINC - interp- v-round0 0.860 usec       iCrc=8108
- PLAINC -           round1 0.857 usec       iCrc=8105
- PLAINC - interp-hv-round0 2.103 usec       iCrc=8112
- PLAINC -           round1 2.050 usec       iCrc=8103
-  ---
- MMX    - interp- h-round0 0.105 usec       iCrc=8107
- MMX    -           round1 0.106 usec       iCrc=8100
- MMX    - interp- v-round0 0.106 usec       iCrc=8108
- MMX    -           round1 0.106 usec       iCrc=8105
- MMX    - interp-hv-round0 0.145 usec       iCrc=8112
- MMX    -           round1 0.145 usec       iCrc=8103
-  ---
- MMXEXT - interp- h-round0 0.028 usec       iCrc=8107
- MMXEXT -           round1 0.041 usec       iCrc=8100
- MMXEXT - interp- v-round0 0.027 usec       iCrc=8108
- MMXEXT -           round1 0.041 usec       iCrc=8105
- MMXEXT - interp-hv-round0 0.066 usec       iCrc=8112
- MMXEXT -           round1 0.065 usec       iCrc=8103
-  ---
- SSE2   - interp- h-round0 0.109 usec       iCrc=8107
- SSE2   -           round1 0.105 usec       iCrc=8100
- SSE2   - interp- v-round0 0.106 usec       iCrc=8108
- SSE2   -           round1 0.109 usec       iCrc=8105
- SSE2   - interp-hv-round0 0.145 usec       iCrc=8112
- SSE2   -           round1 0.145 usec       iCrc=8103
-  ---
-DNOW  - skipped...
-DNOWE - skipped...
-  ======  test SAD ======
- PLAINC - sad8    0.251 usec       sad=3776
- PLAINC - sad16   1.601 usec       sad=27214
- PLAINC - sad16bi 2.371 usec       sad=26274
- PLAINC - dev16   1.564 usec       sad=3344
-  ---
- MMX    - sad8    0.057 usec       sad=3776
- MMX    - sad16   0.182 usec       sad=27214
- MMX    - sad16bi 2.462 usec       sad=26274
- MMX    - dev16   0.311 usec       sad=3344
-  ---
- MMXEXT - sad8    0.036 usec       sad=3776
- MMXEXT - sad16   0.109 usec       sad=27214
- MMXEXT - sad16bi 0.143 usec       sad=26274
- MMXEXT - dev16   0.192 usec       sad=3344
-  ---
- SSE2   - sad8    0.057 usec       sad=3776
- SSE2   - sad16   0.179 usec       sad=27214
- SSE2   - sad16bi 2.456 usec       sad=26274
- SSE2   - dev16   0.321 usec       sad=3344
-  ---
-DNOW  - skipped...
-DNOWE - skipped...
-  ===  test transfer ===
- PLAINC - 8to16     0.151 usec       crc=28288
- PLAINC - 16to8     1.113 usec       crc=28288
- PLAINC - 8to8      0.043 usec       crc=20352
- PLAINC - 16to8add  1.069 usec       crc=25536
- PLAINC - 8to16sub  0.631 usec       crc1=28064 crc2=16256
- PLAINC - 8to16sub2 0.597 usec       crc=20384
-  ---
- MMX    - 8to16     0.032 usec       crc=28288
- MMX    - 16to8     0.024 usec       crc=28288
- MMX    - 8to8      0.020 usec       crc=20352
- MMX    - 16to8add  0.043 usec       crc=25536
- MMX    - 8to16sub  0.066 usec       crc1=28064 crc2=16256
- MMX    - 8to16sub2 0.111 usec       crc=20384
-  ---
-  =====  test quant =====
- PLAINC -   quant4_intra 74.248 usec       crc=29809
- PLAINC -   quant4_inter 70.850 usec       crc=12574
- PLAINC - dequant4_intra 40.628 usec       crc=24052
- PLAINC - dequant4_inter 45.691 usec       crc=63847
- PLAINC -    quant_intra 43.357 usec       crc=25662
- PLAINC -    quant_inter 33.410 usec       crc=23972
- PLAINC -  dequant_intra 36.384 usec       crc=49900
- PLAINC -  dequant_inter 48.930 usec       crc=48899
-  ---
- MMX    -   quant4_intra 7.445 usec       crc=3459
- *** CRC ERROR! ***
- MMX    -   quant4_inter 5.384 usec       crc=51072
- *** CRC ERROR! ***
- MMX    - dequant4_intra 5.515 usec       crc=24052
- MMX    - dequant4_inter 7.745 usec       crc=63847
- MMX    -    quant_intra 4.661 usec       crc=25662
- MMX    -    quant_inter 4.406 usec       crc=23972
- MMX    -  dequant_intra 4.928 usec       crc=49900
- MMX    -  dequant_inter 4.532 usec       crc=48899
-  ---
-  =====  test cbp =====
- PLAINC -   calc_cbp#1 0.371 usec       cbp=0x15
- PLAINC -   calc_cbp#2 0.432 usec       cbp=0x38
- PLAINC -   calc_cbp#3 0.339 usec       cbp=0xf
- PLAINC -   calc_cbp#4 0.506 usec       cbp=0x5
-  ---
- MMX    -   calc_cbp#1 0.136 usec       cbp=0x15
- MMX    -   calc_cbp#2 0.134 usec       cbp=0x38
- MMX    -   calc_cbp#3 0.138 usec       cbp=0xf
- MMX    -   calc_cbp#4 0.135 usec       cbp=0x5
-  ---
- SSE2   -   calc_cbp#1 0.136 usec       cbp=0x15
- SSE2   -   calc_cbp#2 0.133 usec       cbp=0x38
- SSE2   -   calc_cbp#3 0.133 usec       cbp=0xf
- SSE2   -   calc_cbp#4 0.141 usec       cbp=0x5
-  ---
- */
+ /*********************************************************************/

 Legend:



Removed from v.851
 


changed lines


 
Added in v.2021
 Legend:



Removed from v.851
 


changed lines


 
Added in v.2021
-Removed from v.851
+Added in v.2021

No admin address has been configured	ViewVC Help
Powered by ViewVC 1.0.4