"psubw      %%mm3,  %%mm2\n"            "psllw      $2,     %%mm3\n"            "psubw      %%mm3,  %%mm2\n"            "movq       (%%eax,%1),%%mm1\n"  /* load pix */            "movq       %%mm1,  %%mm3\n"            "punpcklbw  %%mm7,  %%mm1\n"            "punpckhbw  %%mm7,  %%mm3\n"            "psllw      $2,     %%mm1\n"            "paddw      %%mm1,  %%mm0\n"            "psllw      $2,     %%mm1\n"            "paddw      %%mm1,  %%mm0\n"            "psllw      $2,     %%mm3\n"            "paddw      %%mm3,  %%mm2\n"            "psllw      $2,     %%mm3\n"            "paddw      %%mm3,  %%mm2\n"            "movq       (%%eax,%1,2),%%mm1\n"  /* load pix+1 */            "movq       %%mm1,  %%mm3\n"            "punpcklbw  %%mm7,  %%mm1\n"            "punpckhbw  %%mm7,  %%mm3\n"            "psllw      $2,     %%mm1\n"            "paddw      %%mm1,  %%mm0\n"            "psllw      $2,     %%mm1\n"            "paddw      %%mm1,  %%mm0\n"            "psllw      $2,     %%mm3\n"            "paddw      %%mm3,  %%mm2\n"            "psllw      $2,     %%mm3\n"            "paddw      %%mm3,  %%mm2\n"            "movq       (%0,%1,4),%%mm1\n"  /* load pix+2 */            "movq       %%mm1,  %%mm3\n"            "punpcklbw  %%mm7,  %%mm1\n"            "punpckhbw  %%mm7,  %%mm3\n"            "psubw      %%mm1,  %%mm0\n"            "psllw      $2,     %%mm1\n"            "psubw      %%mm1,  %%mm0\n"            "psubw      %%mm3,  %%mm2\n"            "psllw      $2,     %%mm3\n"            "psubw      %%mm3,  %%mm2\n"            "movq       (%%eax,%1,4),%%mm1\n"  /* load pix+3 */            "movq       %%mm1,  %%mm3\n"            "punpcklbw  %%mm7,  %%mm1\n"            "punpckhbw  %%mm7,  %%mm3\n"            "paddw      %%mm1,  %%mm0\n"            "paddw      %%mm3,  %%mm2\n"            "movq       %%mm0,   (%2)\n"            "movq       %%mm2,  8(%2)\n"            "addl   $8,         %%eax\n"            "addl   $8,         %0\n"            "movq       (%0),   %%mm0\n"    /* load pix-2 */            "movq       %%mm0,  %%mm2\n"            "punpcklbw  %%mm7,  %%mm0\n"            "punpckhbw  %%mm7,  %%mm2\n"            "movq       (%%eax),%%mm1\n"    /* load pix-1 */            "movq       %%mm1,  %%mm3\n"            "punpcklbw  %%mm7,  %%mm1\n"            "punpckhbw  %%mm7,  %%mm3\n"            "psubw      %%mm1,  %%mm0\n"            "psllw      $2,     %%mm1\n"            "psubw      %%mm1,  %%mm0\n"            "psubw      %%mm3,  %%mm2\n"            "psllw      $2,     %%mm3\n"            "psubw      %%mm3,  %%mm2\n"            "movq       (%%eax,%1),%%mm1\n"  /* load pix */            "movq       %%mm1,  %%mm3\n"            "punpcklbw  %%mm7,  %%mm1\n"            "punpckhbw  %%mm7,  %%mm3\n"            "psllw      $2,     %%mm1\n"            "paddw      %%mm1,  %%mm0\n"            "psllw      $2,     %%mm1\n"            "paddw      %%mm1,  %%mm0\n"            "psllw      $2,     %%mm3\n"            "paddw      %%mm3,  %%mm2\n"            "psllw      $2,     %%mm3\n"            "paddw      %%mm3,  %%mm2\n"            "movq       (%%eax,%1,2),%%mm1\n"  /* load pix+1 */            "movq       %%mm1,  %%mm3\n"            "punpcklbw  %%mm7,  %%mm1\n"            "punpckhbw  %%mm7,  %%mm3\n"            "psllw      $2,     %%mm1\n"            "paddw      %%mm1,  %%mm0\n"            "psllw      $2,     %%mm1\n"            "paddw      %%mm1,  %%mm0\n"            "psllw      $2,     %%mm3\n"            "paddw      %%mm3,  %%mm2\n"            "psllw      $2,     %%mm3\n"            "paddw      %%mm3,  %%mm2\n"            "movq       (%0,%1,4),%%mm1\n"  /* load pix+2 */            "movq       %%mm1,  %%mm3\n"            "punpcklbw  %%mm7,  %%mm1\n"            "punpckhbw  %%mm7,  %%mm3\n"            "psubw      %%mm1,  %%mm0\n"            "psllw      $2,     %%mm1\n"            "psubw      %%mm1,  %%mm0\n"            "psubw      %%mm3,  %%mm2\n"            "psllw      $2,     %%mm3\n"            "psubw      %%mm3,  %%mm2\n"            "movq       (%%eax,%1,4),%%mm1\n"  /* load pix+3 */            "movq       %%mm1,  %%mm3\n"            "punpcklbw  %%mm7,  %%mm1\n"            "punpckhbw  %%mm7,  %%mm3\n"            "paddw      %%mm1,  %%mm0\n"            "paddw      %%mm3,  %%mm2\n"            "movq       %%mm0,  16(%2)\n"            "movq       %%mm2,  24(%2)\n"            "addl   $8,         %%eax\n"            "addl   $8,         %0\n"            "movd       (%0),   %%mm0\n"    /* load pix-2 */            "punpcklbw  %%mm7,  %%mm0\n"            "movd       (%%eax),%%mm1\n"    /* load pix-1 */            "punpcklbw  %%mm7,  %%mm1\n"            "psubw      %%mm1,  %%mm0\n"            "psllw      $2,     %%mm1\n"            "psubw      %%mm1,  %%mm0\n"            "movd       (%%eax,%1),%%mm1\n"  /* load pix */            "punpcklbw  %%mm7,  %%mm1\n"            "psllw      $2,     %%mm1\n"            "paddw      %%mm1,  %%mm0\n"            "psllw      $2,     %%mm1\n"            "paddw      %%mm1,  %%mm0\n"            "movd       (%%eax,%1,2),%%mm1\n"  /* load pix+1 */            "punpcklbw  %%mm7,  %%mm1\n"            "psllw      $2,     %%mm1\n"            "paddw      %%mm1,  %%mm0\n"            "psllw      $2,     %%mm1\n"            "paddw      %%mm1,  %%mm0\n"            "movd       (%0,%1,4),%%mm1\n"  /* load pix+2 */            "punpcklbw  %%mm7,  %%mm1\n"            "psubw      %%mm1,  %%mm0\n"            "psllw      $2,     %%mm1\n"            "psubw      %%mm1,  %%mm0\n"            "movd       (%%eax,%1,4),%%mm1\n"  /* load pix+3 */            "punpcklbw  %%mm7,  %%mm1\n"            "paddw      %%mm1,  %%mm0\n"            "movq       %%mm0,  32(%2)\n"            : : "r"(src-2*i_src_stride-2), "r"(i_src_stride), "r"(&tap[0]) : "%eax" );        /* last one */        tap[16+4] = x264_tapfilter( &src[-2+16+4], i_src_stride );        for( x = 0; x < 16; x++ )        {            dst[x] = x264_mc_clip1( ( tap[0+x] - 5*tap[1+x] + 20 * tap[2+x] + 20 * tap[3+x] -5*tap[4+x] + tap[5+x] + 512 ) >> 10 );        }        src += i_src_stride;        dst += i_dst_stride;    }#if 0#if 0    int i, x, y;    for( y = 0; y < i_height; y++ )    {        int16_t tap[5+16];        for( i = 0; i < 5+16; i++ )        {            tap[i] = x264_tapfilter( &src[-2+i], i_src_stride );        }        for( x = 0; x < 16; x++ )        {            dst[x] = x264_mc_clip1( ( tap[0+x] - 5*tap[1+x] + 20 * tap[2+x] + 20 * tap[3+x] -5*tap[4+x] + tap[5+x] + 512 ) >> 10 );        }        src += i_src_stride;        dst += i_dst_stride;    }#else    mc_hc_w8( src,     i_src_stride, dst,     i_dst_stride, i_height );    mc_hc_w8( &src[8], i_src_stride, &dst[8], i_dst_stride, i_height );#endif#endif}/* mc I+H */static void mc_xy10_w16( uint8_t *src, int i_src_stride, uint8_t *dst, int i_dst_stride, int i_height ){    uint8_t tmp[16*16];    mc_hh_w16( src, i_src_stride, tmp, 16, i_height );    pixel_avg_w16( dst, i_dst_stride, src, i_src_stride, tmp, 16, i_height );}static void mc_xy30_w16( uint8_t *src, int i_src_stride, uint8_t *dst, int i_dst_stride, int i_height ){    uint8_t tmp[16*16];    mc_hh_w16( src, i_src_stride, tmp, 16, i_height );    pixel_avg_w16( dst, i_dst_stride, src+1, i_src_stride, tmp, 16, i_height );}/* mc I+V */static void mc_xy01_w16( uint8_t *src, int i_src_stride, uint8_t *dst, int i_dst_stride, int i_height ){    uint8_t tmp[16*16];    mc_hv_w16( src, i_src_stride, tmp, 16, i_height );    pixel_avg_w16( dst, i_dst_stride, src, i_src_stride, tmp, 16, i_height );}static void mc_xy03_w16( uint8_t *src, int i_src_stride, uint8_t *dst, int i_dst_stride, int i_height ){    uint8_t tmp[16*16];    mc_hv_w16( src, i_src_stride, tmp, 16, i_height );    pixel_avg_w16( dst, i_dst_stride, src+i_src_stride, i_src_stride, tmp, 16, i_height );}/* H+V */static void mc_xy11_w16( uint8_t *src, int i_src_stride, uint8_t *dst, int i_dst_stride, int i_height ){    uint8_t tmp1[16*16];    uint8_t tmp2[16*16];    mc_hv_w16( src, i_src_stride, tmp1, 16, i_height );    mc_hh_w16( src, i_src_stride, tmp2, 16, i_height );    pixel_avg_w16( dst, i_dst_stride, tmp1, 16, tmp2, 16, i_height );}static void mc_xy31_w16( uint8_t *src, int i_src_stride, uint8_t *dst, int i_dst_stride, int i_height ){    uint8_t tmp1[16*16];    uint8_t tmp2[16*16];    mc_hv_w16( src+1, i_src_stride, tmp1, 16, i_height );    mc_hh_w16( src,   i_src_stride, tmp2, 16, i_height );    pixel_avg_w16( dst, i_dst_stride, tmp1, 16, tmp2, 16, i_height );}static void mc_xy13_w16( uint8_t *src, int i_src_stride, uint8_t *dst, int i_dst_stride, int i_height ){    uint8_t tmp1[16*16];    uint8_t tmp2[16*16];    mc_hv_w16( src,              i_src_stride, tmp1, 16, i_height );    mc_hh_w16( src+i_src_stride, i_src_stride, tmp2, 16, i_height );    pixel_avg_w16( dst, i_dst_stride, tmp1, 16, tmp2, 16, i_height );}static void mc_xy33_w16( uint8_t *src, int i_src_stride, uint8_t *dst, int i_dst_stride, int i_height ){    uint8_t tmp1[16*16];    uint8_t tmp2[16*16];    mc_hv_w16( src+1,            i_src_stride, tmp1, 16, i_height );    mc_hh_w16( src+i_src_stride, i_src_stride, tmp2, 16, i_height );    pixel_avg_w16( dst, i_dst_stride, tmp1, 16, tmp2, 16, i_height );}static void mc_xy21_w16( uint8_t *src, int i_src_stride, uint8_t *dst, int i_dst_stride, int i_height ){    uint8_t tmp1[16*16];    uint8_t tmp2[16*16];    mc_hc_w16( src, i_src_stride, tmp1, 16, i_height );    mc_hh_w16( src, i_src_stride, tmp2, 16, i_height );    pixel_avg_w16( dst, i_dst_stride, tmp1, 16, tmp2, 16, i_height );}static void mc_xy12_w16( uint8_t *src, int i_src_stride, uint8_t *dst, int i_dst_stride, int i_height ){    uint8_t tmp1[16*16];    uint8_t tmp2[16*16];    mc_hc_w16( src, i_src_stride, tmp1, 16, i_height );    mc_hv_w16( src, i_src_stride, tmp2, 16, i_height );    pixel_avg_w16( dst, i_dst_stride, tmp1, 16, tmp2, 16, i_height );}static void mc_xy32_w16( uint8_t *src, int i_src_stride, uint8_t *dst, int i_dst_stride, int i_height ){    uint8_t tmp1[16*16];    uint8_t tmp2[16*16];    mc_hc_w16( src,   i_src_stride, tmp1, 16, i_height );    mc_hv_w16( src+1, i_src_stride, tmp2, 16, i_height );    pixel_avg_w16( dst, i_dst_stride, tmp1, 16, tmp2, 16, i_height );}static void mc_xy23_w16( uint8_t *src, int i_src_stride, uint8_t *dst, int i_dst_stride, int i_height ){    uint8_t tmp1[16*16];    uint8_t tmp2[16*16];    mc_hc_w16( src,              i_src_stride, tmp1, 16, i_height );    mc_hh_w16( src+i_src_stride, i_src_stride, tmp2, 16, i_height );    pixel_avg_w16( dst, i_dst_stride, tmp1, 16, tmp2, 16, i_height );}static void motion_compensation_luma( uint8_t *src, int i_src_stride,                                      uint8_t *dst, int i_dst_stride,                                      int mvx,int mvy,                                      int i_width, int i_height ){    static const pf_mc_t pf_mc[3][4][4] =    /*XXX [dqy][dqx] */    {        {            { mc_copy_w4,  mc_xy10_w4,    mc_hh_w4,      mc_xy30_w4 },            { mc_xy01_w4,  mc_xy11_w4,    mc_xy21_w4,    mc_xy31_w4 },            { mc_hv_w4,    mc_xy12_w4,    mc_hc_w4,      mc_xy32_w4 },            { mc_xy03_w4,  mc_xy13_w4,    mc_xy23_w4,    mc_xy33_w4 },        },        {            { mc_copy_w8,  mc_xy10_w8,    mc_hh_w8,      mc_xy30_w8 },            { mc_xy01_w8,  mc_xy11_w8,    mc_xy21_w8,    mc_xy31_w8 },            { mc_hv_w8,    mc_xy12_w8,    mc_hc_w8,      mc_xy32_w8 },            { mc_xy03_w8,  mc_xy13_w8,    mc_xy23_w8,    mc_xy33_w8 },        },        {            { mc_copy_w16,  mc_xy10_w16,    mc_hh_w16,      mc_xy30_w16 },            { mc_xy01_w16,  mc_xy11_w16,    mc_xy21_w16,    mc_xy31_w16 },            { mc_hv_w16,    mc_xy12_w16,    mc_hc_w16,      mc_xy32_w16 },            { mc_xy03_w16,  mc_xy13_w16,    mc_xy23_w16,    mc_xy33_w16 },        }    };    src += (mvy >> 2) * i_src_stride + (mvx >> 2);    if( i_width == 4 )    {        pf_mc[0][mvy&0x03][mvx&0x03]( src, i_src_stride, dst, i_dst_stride, i_height );    }    else if( i_width == 8 )    {        pf_mc[1][mvy&0x03][mvx&0x03]( src, i_src_stride, dst, i_dst_stride, i_height );    }    else if( i_width == 16 )    {        pf_mc[2][mvy&0x03][mvx&0x03]( src, i_src_stride, dst, i_dst_stride, i_height );    }    else    {        fprintf( stderr, "Error: motion_compensation_luma called with invalid width" );    }}#if 0/* full chroma mc (ie until 1/8 pixel)*/static void motion_compensation_chroma( uint8_t *src, int i_src_stride,                                        uint8_t *dst, int i_dst_stride,                                        int mvx, int mvy,                                        int i_width, int i_height ){    int x, y;    int d8x = mvx&0x07;    int d8y = mvy&0x07;    int cA = (8-d8x)*(8-d8y);    int cB = d8x    *(8-d8y);    int cC = (8-d8x)*d8y;    int cD = d8x    *d8y;    src += (mvy >> 3) * i_dst_stride + (mvx >> 3);    for( y = 0; y < i_height; y++ )    {        for( x = 0; x < i_width; x++ )        {            dst[x] = ( cA*src[x]              + cB*src[x+1] +                       cC*src[i_dst_stride+x] + cD*src[i_dst_stride+x+1] + 32 ) >> 6;        }        dst += i_src_stride;        src += i_dst_stride;    }}#endifvoid x264_mc_mmxext_init( x264_mc_t pf[2] ){    pf[MC_LUMA]   = motion_compensation_luma;#if 0    pf[MC_CHROMA] = motion_compensation_chroma;#endif}