mc-c.c.svn-base

H.264 source codes

MC_CH( mc_xy21, mmxext, 4, 8, 0 )
MC_CH( mc_xy23, mmxext, 4, 8, i_src_stride )

MC_CV( mc_xy12, mmxext, 4, 8, 0 )
MC_CV( mc_xy32, mmxext, 4, 8, 1 )

#if 0
static void mc_xy10_w4( uint8_t *src, int i_src_stride, uint8_t *dst, int i_dst_stride, int i_height )
{
    uint8_t tmp[4*8];
    mc_hh_w4( src, i_src_stride, tmp, 4, i_height );
    pixel_avg_w4( dst, i_dst_stride, src, i_src_stride, tmp, 4, i_height );
}
static void mc_xy30_w4( uint8_t *src, int i_src_stride, uint8_t *dst, int i_dst_stride, int i_height )
{
    uint8_t tmp[4*8];
    mc_hh_w4( src, i_src_stride, tmp, 4, i_height );
    pixel_avg_w4( dst, i_dst_stride, src+1, i_src_stride, tmp, 4, i_height );
}

static void mc_xy01_w4( uint8_t *src, int i_src_stride, uint8_t *dst, int i_dst_stride, int i_height )
{
    uint8_t tmp[4*8];
    mc_hv_w4( src, i_src_stride, tmp, 4, i_height );
    pixel_avg_w4( dst, i_dst_stride, src, i_src_stride, tmp, 4, i_height );
}
static void mc_xy03_w4( uint8_t *src, int i_src_stride, uint8_t *dst, int i_dst_stride, int i_height )
{
    uint8_t tmp[4*8];
    mc_hv_w4( src, i_src_stride, tmp, 4, i_height );
    pixel_avg_w4( dst, i_dst_stride, src+i_src_stride, i_src_stride, tmp, 4, i_height );
}

static void mc_xy11_w4( uint8_t *src, int i_src_stride, uint8_t *dst, int i_dst_stride, int i_height )
{
    uint8_t tmp1[4*8];
    uint8_t tmp2[4*8];

    mc_hv_w4( src, i_src_stride, tmp1, 4, i_height );
    mc_hh_w4( src, i_src_stride, tmp2, 4, i_height );
    pixel_avg_w4( dst, i_dst_stride, tmp1, 4, tmp2, 4, i_height );
}
static void mc_xy31_w4( uint8_t *src, int i_src_stride, uint8_t *dst, int i_dst_stride, int i_height )
{
    uint8_t tmp1[4*8];
    uint8_t tmp2[4*8];

    mc_hv_w4( src+1, i_src_stride, tmp1, 4, i_height );
    mc_hh_w4( src,   i_src_stride, tmp2, 4, i_height );
    pixel_avg_w4( dst, i_dst_stride, tmp1, 4, tmp2, 4, i_height );
}
static void mc_xy13_w4( uint8_t *src, int i_src_stride, uint8_t *dst, int i_dst_stride, int i_height )
{
    uint8_t tmp1[4*8];
    uint8_t tmp2[4*8];

    mc_hv_w4( src,              i_src_stride, tmp1, 4, i_height );
    mc_hh_w4( src+i_src_stride, i_src_stride, tmp2, 4, i_height );
    pixel_avg_w4( dst, i_dst_stride, tmp1, 4, tmp2, 4, i_height );
}
static void mc_xy33_w4( uint8_t *src, int i_src_stride, uint8_t *dst, int i_dst_stride, int i_height )
{
    uint8_t tmp1[4*8];
    uint8_t tmp2[4*8];

    mc_hv_w4( src+1,            i_src_stride, tmp1, 4, i_height );
    mc_hh_w4( src+i_src_stride, i_src_stride, tmp2, 4, i_height );
    pixel_avg_w4( dst, i_dst_stride, tmp1, 4, tmp2, 4, i_height );
}

static void mc_xy21_w4( uint8_t *src, int i_src_stride, uint8_t *dst, int i_dst_stride, int i_height )
{
    uint8_t tmp1[4*8];
    uint8_t tmp2[4*8];

    mc_hc_w4( src, i_src_stride, tmp1, 4, i_height );
    mc_hh_w4( src, i_src_stride, tmp2, 4, i_height );
    pixel_avg_w4( dst, i_dst_stride, tmp1, 4, tmp2, 4, i_height );
}
static void mc_xy23_w4( uint8_t *src, int i_src_stride, uint8_t *dst, int i_dst_stride, int i_height )
{
    uint8_t tmp1[4*8];
    uint8_t tmp2[4*8];

    mc_hc_w4( src,              i_src_stride, tmp1, 4, i_height );
    mc_hh_w4( src+i_src_stride, i_src_stride, tmp2, 4, i_height );
    pixel_avg_w4( dst, i_dst_stride, tmp1, 4, tmp2, 4, i_height );
}

static void mc_xy12_w4( uint8_t *src, int i_src_stride, uint8_t *dst, int i_dst_stride, int i_height )
{
    uint8_t tmp1[4*8];
    uint8_t tmp2[4*8];

    mc_hc_w4( src, i_src_stride, tmp1, 4, i_height );
    mc_hv_w4( src, i_src_stride, tmp2, 4, i_height );
    pixel_avg_w4( dst, i_dst_stride, tmp1, 4, tmp2, 4, i_height );
}
static void mc_xy32_w4( uint8_t *src, int i_src_stride, uint8_t *dst, int i_dst_stride, int i_height )
{
    uint8_t tmp1[4*8];
    uint8_t tmp2[4*8];

    mc_hc_w4( src,   i_src_stride, tmp1, 4, i_height );
    mc_hv_w4( src+1, i_src_stride, tmp2, 4, i_height );
    pixel_avg_w4( dst, i_dst_stride, tmp1, 4, tmp2, 4, i_height );
}
#endif

/*****************************************************************************
 * MC with width == 8 (height <= 16)
 *****************************************************************************/

static inline void mc_hh_w8( uint8_t *src, int i_src, uint8_t *dst, int i_dst, int i_height )
{
    mc_hh_w4( &src[0], i_src, &dst[0], i_dst, i_height );
    mc_hh_w4( &src[4], i_src, &dst[4], i_dst, i_height );
}
static inline void mc_hv_w8( uint8_t *src, int i_src, uint8_t *dst, int i_dst, int i_height )
{
    int y;

    src -= 2 * i_src;

    MMX_ZERO( %%mm7 );
    MMX_INIT( %%mm6, x264_w0x10 );

    for( y = 0; y < i_height; y++ )
    {
        MMX_LOAD_2x8( %%mm0, %%mm5, %%mm1, %%mm2, %%mm7,  &src[0*i_src], i_src );
        MMX_FILTERTAP2_P1( %%mm0, %%mm1, %%mm5, %%mm2 );


        MMX_LOAD_2x8( %%mm1, %%mm3, %%mm2, %%mm4, %%mm7,  &src[2*i_src], i_src );
        MMX_FILTERTAP2_P2( %%mm0, %%mm1, %%mm2, %%mm5, %%mm3, %%mm4 );

        MMX_LOAD_2x8( %%mm1, %%mm3, %%mm2, %%mm4, %%mm7,  &src[4*i_src], i_src );
        MMX_FILTERTAP_P3( %%mm0, %%mm1, %%mm2, %%mm6, %%mm7 );
        MMX_FILTERTAP_P3( %%mm5, %%mm3, %%mm4, %%mm6, %%mm7 );

        MMX_SAVEPACK_8P( %%mm0, %%mm5, %%mm7, dst );

        src += i_src;
        dst += i_dst;
    }
}

static inline void mc_hc_w8( uint8_t *src, int i_src_stride, uint8_t *dst, int i_dst_stride, int i_height )
{
    int x, y;

    asm volatile( "pxor %%mm7,        %%mm7\n" : : );

    for( y = 0; y < i_height; y++ )
    {
        int16_t tap[5+8];

        /* first 8 */
        asm volatile(
            "leal   (%0, %1),   %%eax\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,   (%2)\n"
            "movq       %%mm2,  8(%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,  16(%2)\n"
            : : "r"(src-2*i_src_stride-2), "r"(i_src_stride), "r"(&tap[0]) : "%eax" );

        /* last one */
        tap[8+4] = x264_tapfilter( &src[-2+8+4], i_src_stride );

        for( x = 0; x < 8; 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;
    }
}

MC_IH( mc_xy10, mmxext, 8, 16, 0 )
MC_IH( mc_xy30, mmxext, 8, 16, 1 )

MC_IV( mc_xy01, mmxext, 8, 16, 0 )
MC_IV( mc_xy03, mmxext, 8, 16, i_src_stride )

MC_HV( mc_xy11, mmxext, 8, 16, 0, 0 )
MC_HV( mc_xy31, mmxext, 8, 16, 1, 0 )
MC_HV( mc_xy13, mmxext, 8, 16, 0, i_src_stride )
MC_HV( mc_xy33, mmxext, 8, 16, 1, i_src_stride )

MC_CH( mc_xy21, mmxext, 8, 16, 0 )
MC_CH( mc_xy23, mmxext, 8, 16, i_src_stride )

MC_CV( mc_xy12, mmxext, 8, 16, 0 )
MC_CV( mc_xy32, mmxext, 8, 16, 1 )

#if 0
/* mc I+H */
static void mc_xy10_w8( uint8_t *src, int i_src_stride, uint8_t *dst, int i_dst_stride, int i_height )
{
    uint8_t tmp[8*16];
    mc_hh_w8( src, i_src_stride, tmp, 8, i_height );
    pixel_avg_w8( dst, i_dst_stride, src, i_src_stride, tmp, 8, i_height );
}
static void mc_xy30_w8( uint8_t *src, int i_src_stride, uint8_t *dst, int i_dst_stride, int i_height )
{
    uint8_t tmp[8*16];
    mc_hh_w8( src, i_src_stride, tmp, 8, i_height );
    pixel_avg_w8( dst, i_dst_stride, src+1, i_src_stride, tmp, 8, i_height );
}
/* mc I+V */
static void mc_xy01_w8( uint8_t *src, int i_src_stride, uint8_t *dst, int i_dst_stride, int i_height )
{
    uint8_t tmp[8*16];
    mc_hv_w8( src, i_src_stride, tmp, 8, i_height );
    pixel_avg_w8( dst, i_dst_stride, src, i_src_stride, tmp, 8, i_height );
}
static void mc_xy03_w8( uint8_t *src, int i_src_stride, uint8_t *dst, int i_dst_stride, int i_height )
{
    uint8_t tmp[8*16];
    mc_hv_w8( src, i_src_stride, tmp, 8, i_height );
    pixel_avg_w8( dst, i_dst_stride, src+i_src_stride, i_src_stride, tmp, 8, i_height );
}
/* H+V */
static void mc_xy11_w8( uint8_t *src, int i_src_stride, uint8_t *dst, int i_dst_stride, int i_height )
{
    uint8_t tmp1[8*16];
    uint8_t tmp2[8*16];

    mc_hv_w8( src, i_src_stride, tmp1, 8, i_height );
    mc_hh_w8( src, i_src_stride, tmp2, 8, i_height );
    pixel_avg_w8( dst, i_dst_stride, tmp1, 8, tmp2, 8, i_height );
}
static void mc_xy31_w8( uint8_t *src, int i_src_stride, uint8_t *dst, int i_dst_stride, int i_height )
{
    uint8_t tmp1[8*16];
    uint8_t tmp2[8*16];

    mc_hv_w8( src+1, i_src_stride, tmp1, 8, i_height );
    mc_hh_w8( src,   i_src_stride, tmp2, 8, i_height );
    pixel_avg_w8( dst, i_dst_stride, tmp1, 8, tmp2, 8, i_height );
}
static void mc_xy13_w8( uint8_t *src, int i_src_stride, uint8_t *dst, int i_dst_stride, int i_height )
{
    uint8_t tmp1[8*16];
    uint8_t tmp2[8*16];

    mc_hv_w8( src,              i_src_stride, tmp1, 8, i_height );
    mc_hh_w8( src+i_src_stride, i_src_stride, tmp2, 8, i_height );
    pixel_avg_w8( dst, i_dst_stride, tmp1, 8, tmp2, 8, i_height );
}
static void mc_xy33_w8( uint8_t *src, int i_src_stride, uint8_t *dst, int i_dst_stride, int i_height )
{
    uint8_t tmp1[8*16];
    uint8_t tmp2[8*16];

    mc_hv_w8( src+1,            i_src_stride, tmp1, 8, i_height );
    mc_hh_w8( src+i_src_stride, i_src_stride, tmp2, 8, i_height );
    pixel_avg_w8( dst, i_dst_stride, tmp1, 8, tmp2, 8, i_height );
}
static void mc_xy21_w8( uint8_t *src, int i_src_stride, uint8_t *dst, int i_dst_stride, int i_height )
{
    uint8_t tmp1[8*16];
    uint8_t tmp2[8*16];

    mc_hc_w8( src, i_src_stride, tmp1, 8, i_height );
    mc_hh_w8( src, i_src_stride, tmp2, 8, i_height );
    pixel_avg_w8( dst, i_dst_stride, tmp1, 8, tmp2, 8, i_height );
}
static void mc_xy12_w8( uint8_t *src, int i_src_stride, uint8_t *dst, int i_dst_stride, int i_height )
{
    uint8_t tmp1[8*16];
    uint8_t tmp2[8*16];

    mc_hc_w8( src, i_src_stride, tmp1, 8, i_height );
    mc_hv_w8( src, i_src_stride, tmp2, 8, i_height );
    pixel_avg_w8( dst, i_dst_stride, tmp1, 8, tmp2, 8, i_height );
}
static void mc_xy32_w8( uint8_t *src, int i_src_stride, uint8_t *dst, int i_dst_stride, int i_height )
{
    uint8_t tmp1[8*16];
    uint8_t tmp2[8*16];

    mc_hc_w8( src,   i_src_stride, tmp1, 8, i_height );
    mc_hv_w8( src+1, i_src_stride, tmp2, 8, i_height );