pixel.c

从服务器上下的x264编码器C源码……希望对大家有帮助……这个是09年4月的

    VEC_HADAMAR( diff0v, diff1v, diff2v, diff3v,
                 temp0v, temp1v, temp2v, temp3v );
    VEC_HADAMAR( diff4v, diff5v, diff6v, diff7v,
                 temp4v, temp5v, temp6v, temp7v );
    VEC_ADD_ABS( temp0v, zero_s32v, satdv );
    VEC_ADD_ABS( temp1v, satdv,     satdv );
    VEC_ADD_ABS( temp2v, satdv,     satdv );
    VEC_ADD_ABS( temp3v, satdv,     satdv );
    VEC_ADD_ABS( temp4v, satdv,     satdv );
    VEC_ADD_ABS( temp5v, satdv,     satdv );
    VEC_ADD_ABS( temp6v, satdv,     satdv );
    VEC_ADD_ABS( temp7v, satdv,     satdv );

    VEC_DIFF_H( pix1, i_pix1, pix2, i_pix2, 8, diff0v, offset1v );
    VEC_DIFF_H( pix1, i_pix1, pix2, i_pix2, 8, diff1v, offset2v );
    VEC_DIFF_H( pix1, i_pix1, pix2, i_pix2, 8, diff2v, offset1v );
    VEC_DIFF_H( pix1, i_pix1, pix2, i_pix2, 8, diff3v, offset2v );
    VEC_DIFF_H( pix1, i_pix1, pix2, i_pix2, 8, diff4v, offset1v );
    VEC_DIFF_H( pix1, i_pix1, pix2, i_pix2, 8, diff5v, offset2v );
    VEC_DIFF_H( pix1, i_pix1, pix2, i_pix2, 8, diff6v, offset1v );
    VEC_DIFF_H( pix1, i_pix1, pix2, i_pix2, 8, diff7v, offset2v );
    VEC_HADAMAR( diff0v, diff1v, diff2v, diff3v,
                 temp0v, temp1v, temp2v, temp3v );
    VEC_HADAMAR( diff4v, diff5v, diff6v, diff7v,
                 temp4v, temp5v, temp6v, temp7v );
    VEC_TRANSPOSE_8( temp0v, temp1v, temp2v, temp3v,
                     temp4v, temp5v, temp6v, temp7v,
                     diff0v, diff1v, diff2v, diff3v,
                     diff4v, diff5v, diff6v, diff7v );
    VEC_HADAMAR( diff0v, diff1v, diff2v, diff3v,
                 temp0v, temp1v, temp2v, temp3v );
    VEC_HADAMAR( diff4v, diff5v, diff6v, diff7v,
                 temp4v, temp5v, temp6v, temp7v );
    VEC_ADD_ABS( temp0v, satdv,     satdv );
    VEC_ADD_ABS( temp1v, satdv,     satdv );
    VEC_ADD_ABS( temp2v, satdv,     satdv );
    VEC_ADD_ABS( temp3v, satdv,     satdv );
    VEC_ADD_ABS( temp4v, satdv,     satdv );
    VEC_ADD_ABS( temp5v, satdv,     satdv );
    VEC_ADD_ABS( temp6v, satdv,     satdv );
    VEC_ADD_ABS( temp7v, satdv,     satdv );

    satdv = vec_sums( satdv, zero_s32v );
    satdv = vec_splat( satdv, 3 );
    vec_ste( satdv, 0, &i_satd );

    return i_satd / 2;
}

/***********************************************************************
 * SATD 16x8
 **********************************************************************/
static int pixel_satd_16x8_altivec( uint8_t *pix1, int i_pix1,
                                    uint8_t *pix2, int i_pix2 )
{
    DECLARE_ALIGNED_16( int i_satd );

    LOAD_ZERO;
    PREP_LOAD;
    PREP_LOAD_SRC( pix2 );
    vec_s32_t satdv;
    vec_s16_t pix1v, pix2v;
    vec_s16_t diffh0v, diffh1v, diffh2v, diffh3v,
              diffh4v, diffh5v, diffh6v, diffh7v;
    vec_s16_t diffl0v, diffl1v, diffl2v, diffl3v,
              diffl4v, diffl5v, diffl6v, diffl7v;
    vec_s16_t temp0v, temp1v, temp2v, temp3v,
              temp4v, temp5v, temp6v, temp7v;

    VEC_DIFF_HL( pix1, i_pix1, pix2, i_pix2, diffh0v, diffl0v );
    VEC_DIFF_HL( pix1, i_pix1, pix2, i_pix2, diffh1v, diffl1v );
    VEC_DIFF_HL( pix1, i_pix1, pix2, i_pix2, diffh2v, diffl2v );
    VEC_DIFF_HL( pix1, i_pix1, pix2, i_pix2, diffh3v, diffl3v );
    VEC_DIFF_HL( pix1, i_pix1, pix2, i_pix2, diffh4v, diffl4v );
    VEC_DIFF_HL( pix1, i_pix1, pix2, i_pix2, diffh5v, diffl5v );
    VEC_DIFF_HL( pix1, i_pix1, pix2, i_pix2, diffh6v, diffl6v );
    VEC_DIFF_HL( pix1, i_pix1, pix2, i_pix2, diffh7v, diffl7v );

    VEC_HADAMAR( diffh0v, diffh1v, diffh2v, diffh3v,
                 temp0v, temp1v, temp2v, temp3v );
    VEC_HADAMAR( diffh4v, diffh5v, diffh6v, diffh7v,
                 temp4v, temp5v, temp6v, temp7v );

    VEC_TRANSPOSE_8( temp0v, temp1v, temp2v, temp3v,
                     temp4v, temp5v, temp6v, temp7v,
                     diffh0v, diffh1v, diffh2v, diffh3v,
                     diffh4v, diffh5v, diffh6v, diffh7v );

    VEC_HADAMAR( diffh0v, diffh1v, diffh2v, diffh3v,
                 temp0v, temp1v, temp2v, temp3v );
    VEC_HADAMAR( diffh4v, diffh5v, diffh6v, diffh7v,
                 temp4v, temp5v, temp6v, temp7v );

    VEC_ADD_ABS( temp0v, zero_s32v, satdv );
    VEC_ADD_ABS( temp1v, satdv,     satdv );
    VEC_ADD_ABS( temp2v, satdv,     satdv );
    VEC_ADD_ABS( temp3v, satdv,     satdv );
    VEC_ADD_ABS( temp4v, satdv,     satdv );
    VEC_ADD_ABS( temp5v, satdv,     satdv );
    VEC_ADD_ABS( temp6v, satdv,     satdv );
    VEC_ADD_ABS( temp7v, satdv,     satdv );

    VEC_HADAMAR( diffl0v, diffl1v, diffl2v, diffl3v,
                 temp0v, temp1v, temp2v, temp3v );
    VEC_HADAMAR( diffl4v, diffl5v, diffl6v, diffl7v,
                 temp4v, temp5v, temp6v, temp7v );

    VEC_TRANSPOSE_8( temp0v, temp1v, temp2v, temp3v,
                     temp4v, temp5v, temp6v, temp7v,
                     diffl0v, diffl1v, diffl2v, diffl3v,
                     diffl4v, diffl5v, diffl6v, diffl7v );

    VEC_HADAMAR( diffl0v, diffl1v, diffl2v, diffl3v,
                 temp0v, temp1v, temp2v, temp3v );
    VEC_HADAMAR( diffl4v, diffl5v, diffl6v, diffl7v,
                 temp4v, temp5v, temp6v, temp7v );

    VEC_ADD_ABS( temp0v, satdv,     satdv );
    VEC_ADD_ABS( temp1v, satdv,     satdv );
    VEC_ADD_ABS( temp2v, satdv,     satdv );
    VEC_ADD_ABS( temp3v, satdv,     satdv );
    VEC_ADD_ABS( temp4v, satdv,     satdv );
    VEC_ADD_ABS( temp5v, satdv,     satdv );
    VEC_ADD_ABS( temp6v, satdv,     satdv );
    VEC_ADD_ABS( temp7v, satdv,     satdv );

    satdv = vec_sums( satdv, zero_s32v );
    satdv = vec_splat( satdv, 3 );
    vec_ste( satdv, 0, &i_satd );

    return i_satd / 2;
}

/***********************************************************************
 * SATD 16x16
 **********************************************************************/
static int pixel_satd_16x16_altivec( uint8_t *pix1, int i_pix1,
                                     uint8_t *pix2, int i_pix2 )
{
    DECLARE_ALIGNED_16( int i_satd );

    LOAD_ZERO;
    PREP_LOAD;
    vec_s32_t satdv;
    vec_s16_t pix1v, pix2v;
    vec_s16_t diffh0v, diffh1v, diffh2v, diffh3v,
              diffh4v, diffh5v, diffh6v, diffh7v;
    vec_s16_t diffl0v, diffl1v, diffl2v, diffl3v,
              diffl4v, diffl5v, diffl6v, diffl7v;
    vec_s16_t temp0v, temp1v, temp2v, temp3v,
              temp4v, temp5v, temp6v, temp7v;
    PREP_LOAD_SRC( pix2 );


    VEC_DIFF_HL( pix1, i_pix1, pix2, i_pix2, diffh0v, diffl0v );
    VEC_DIFF_HL( pix1, i_pix1, pix2, i_pix2, diffh1v, diffl1v );
    VEC_DIFF_HL( pix1, i_pix1, pix2, i_pix2, diffh2v, diffl2v );
    VEC_DIFF_HL( pix1, i_pix1, pix2, i_pix2, diffh3v, diffl3v );
    VEC_DIFF_HL( pix1, i_pix1, pix2, i_pix2, diffh4v, diffl4v );
    VEC_DIFF_HL( pix1, i_pix1, pix2, i_pix2, diffh5v, diffl5v );
    VEC_DIFF_HL( pix1, i_pix1, pix2, i_pix2, diffh6v, diffl6v );
    VEC_DIFF_HL( pix1, i_pix1, pix2, i_pix2, diffh7v, diffl7v );
    VEC_HADAMAR( diffh0v, diffh1v, diffh2v, diffh3v,
                 temp0v, temp1v, temp2v, temp3v );
    VEC_HADAMAR( diffh4v, diffh5v, diffh6v, diffh7v,
                 temp4v, temp5v, temp6v, temp7v );
    VEC_TRANSPOSE_8( temp0v, temp1v, temp2v, temp3v,
                     temp4v, temp5v, temp6v, temp7v,
                     diffh0v, diffh1v, diffh2v, diffh3v,
                     diffh4v, diffh5v, diffh6v, diffh7v );
    VEC_HADAMAR( diffh0v, diffh1v, diffh2v, diffh3v,
                 temp0v, temp1v, temp2v, temp3v );
    VEC_HADAMAR( diffh4v, diffh5v, diffh6v, diffh7v,
                 temp4v, temp5v, temp6v, temp7v );
    VEC_ADD_ABS( temp0v, zero_s32v, satdv );
    VEC_ADD_ABS( temp1v, satdv,     satdv );
    VEC_ADD_ABS( temp2v, satdv,     satdv );
    VEC_ADD_ABS( temp3v, satdv,     satdv );
    VEC_ADD_ABS( temp4v, satdv,     satdv );
    VEC_ADD_ABS( temp5v, satdv,     satdv );
    VEC_ADD_ABS( temp6v, satdv,     satdv );
    VEC_ADD_ABS( temp7v, satdv,     satdv );
    VEC_HADAMAR( diffl0v, diffl1v, diffl2v, diffl3v,
                 temp0v, temp1v, temp2v, temp3v );
    VEC_HADAMAR( diffl4v, diffl5v, diffl6v, diffl7v,
                 temp4v, temp5v, temp6v, temp7v );
    VEC_TRANSPOSE_8( temp0v, temp1v, temp2v, temp3v,
                     temp4v, temp5v, temp6v, temp7v,
                     diffl0v, diffl1v, diffl2v, diffl3v,
                     diffl4v, diffl5v, diffl6v, diffl7v );
    VEC_HADAMAR( diffl0v, diffl1v, diffl2v, diffl3v,
                 temp0v, temp1v, temp2v, temp3v );
    VEC_HADAMAR( diffl4v, diffl5v, diffl6v, diffl7v,
                 temp4v, temp5v, temp6v, temp7v );
    VEC_ADD_ABS( temp0v, satdv,     satdv );
    VEC_ADD_ABS( temp1v, satdv,     satdv );
    VEC_ADD_ABS( temp2v, satdv,     satdv );
    VEC_ADD_ABS( temp3v, satdv,     satdv );
    VEC_ADD_ABS( temp4v, satdv,     satdv );
    VEC_ADD_ABS( temp5v, satdv,     satdv );
    VEC_ADD_ABS( temp6v, satdv,     satdv );
    VEC_ADD_ABS( temp7v, satdv,     satdv );

    VEC_DIFF_HL( pix1, i_pix1, pix2, i_pix2, diffh0v, diffl0v );
    VEC_DIFF_HL( pix1, i_pix1, pix2, i_pix2, diffh1v, diffl1v );
    VEC_DIFF_HL( pix1, i_pix1, pix2, i_pix2, diffh2v, diffl2v );
    VEC_DIFF_HL( pix1, i_pix1, pix2, i_pix2, diffh3v, diffl3v );
    VEC_DIFF_HL( pix1, i_pix1, pix2, i_pix2, diffh4v, diffl4v );
    VEC_DIFF_HL( pix1, i_pix1, pix2, i_pix2, diffh5v, diffl5v );
    VEC_DIFF_HL( pix1, i_pix1, pix2, i_pix2, diffh6v, diffl6v );
    VEC_DIFF_HL( pix1, i_pix1, pix2, i_pix2, diffh7v, diffl7v );
    VEC_HADAMAR( diffh0v, diffh1v, diffh2v, diffh3v,
                 temp0v, temp1v, temp2v, temp3v );
    VEC_HADAMAR( diffh4v, diffh5v, diffh6v, diffh7v,
                 temp4v, temp5v, temp6v, temp7v );
    VEC_TRANSPOSE_8( temp0v, temp1v, temp2v, temp3v,
                     temp4v, temp5v, temp6v, temp7v,
                     diffh0v, diffh1v, diffh2v, diffh3v,
                     diffh4v, diffh5v, diffh6v, diffh7v );
    VEC_HADAMAR( diffh0v, diffh1v, diffh2v, diffh3v,
                 temp0v, temp1v, temp2v, temp3v );
    VEC_HADAMAR( diffh4v, diffh5v, diffh6v, diffh7v,
                 temp4v, temp5v, temp6v, temp7v );
    VEC_ADD_ABS( temp0v, satdv,     satdv );
    VEC_ADD_ABS( temp1v, satdv,     satdv );
    VEC_ADD_ABS( temp2v, satdv,     satdv );
    VEC_ADD_ABS( temp3v, satdv,     satdv );
    VEC_ADD_ABS( temp4v, satdv,     satdv );
    VEC_ADD_ABS( temp5v, satdv,     satdv );
    VEC_ADD_ABS( temp6v, satdv,     satdv );
    VEC_ADD_ABS( temp7v, satdv,     satdv );
    VEC_HADAMAR( diffl0v, diffl1v, diffl2v, diffl3v,
                 temp0v, temp1v, temp2v, temp3v );
    VEC_HADAMAR( diffl4v, diffl5v, diffl6v, diffl7v,
                 temp4v, temp5v, temp6v, temp7v );
    VEC_TRANSPOSE_8( temp0v, temp1v, temp2v, temp3v,
                     temp4v, temp5v, temp6v, temp7v,
                     diffl0v, diffl1v, diffl2v, diffl3v,
                     diffl4v, diffl5v, diffl6v, diffl7v );
    VEC_HADAMAR( diffl0v, diffl1v, diffl2v, diffl3v,
                 temp0v, temp1v, temp2v, temp3v );
    VEC_HADAMAR( diffl4v, diffl5v, diffl6v, diffl7v,
                 temp4v, temp5v, temp6v, temp7v );
    VEC_ADD_ABS( temp0v, satdv,     satdv );
    VEC_ADD_ABS( temp1v, satdv,     satdv );
    VEC_ADD_ABS( temp2v, satdv,     satdv );
    VEC_ADD_ABS( temp3v, satdv,     satdv );
    VEC_ADD_ABS( temp4v, satdv,     satdv );
    VEC_ADD_ABS( temp5v, satdv,     satdv );
    VEC_ADD_ABS( temp6v, satdv,     satdv );
    VEC_ADD_ABS( temp7v, satdv,     satdv );

    satdv = vec_sums( satdv, zero_s32v );
    satdv = vec_splat( satdv, 3 );
    vec_ste( satdv, 0, &i_satd );

    return i_satd / 2;
}



/***********************************************************************
* Interleaved SAD routines
**********************************************************************/

static void pixel_sad_x4_16x16_altivec( uint8_t *fenc,
                                        uint8_t *pix0, uint8_t *pix1,
                                        uint8_t *pix2, uint8_t *pix3,
                                        int i_stride, int scores[4] )
{
    DECLARE_ALIGNED_16( int sum0 );
    DECLARE_ALIGNED_16( int sum1 );
    DECLARE_ALIGNED_16( int sum2 );
    DECLARE_ALIGNED_16( int sum3 );
    int y;

    LOAD_ZERO;
    vec_u8_t temp_lv, temp_hv;
    vec_u8_t fencv, pix0v, pix1v, pix2v, pix3v;
    //vec_u8_t perm0v, perm1v, perm2v, perm3v;
    vec_u8_t perm0vA, perm1vA, perm2vA, perm3vA, perm0vB, perm1vB, perm2vB, perm3vB;

    vec_s32_t sum0v, sum1v, sum2v, sum3v;

    sum0v = vec_splat_s32(0);
    sum1v = vec_splat_s32(0);
    sum2v = vec_splat_s32(0);
    sum3v = vec_splat_s32(0);

    perm0vA = vec_lvsl(0, pix0);
    perm1vA = vec_lvsl(0, pix1);
    perm2vA = vec_lvsl(0, pix2);
    perm3vA = vec_lvsl(0, pix3);

    perm0vB = vec_lvsl(0, pix0 + i_stride);
    perm1vB = vec_lvsl(0, pix1 + i_stride);
    perm2vB = vec_lvsl(0, pix2 + i_stride);
    perm3vB = vec_lvsl(0, pix3 + i_stride);

    for (y = 0; y < 8; y++)
    {
        temp_lv = vec_ld(0, pix0);
        temp_hv = vec_ld(16, pix0);
        pix0v = vec_perm(temp_lv, temp_hv, perm0vA);
        pix0 += i_stride;

        temp_lv = vec_ld(0, pix1);
        temp_hv = vec_ld(16, pix1);
        pix1v = vec_perm(temp_lv, temp_hv, perm1vA);
        pix1 += i_stride;

        fencv = vec_ld(0, fenc);
        fenc += FENC_STRIDE;

        temp_lv = vec_ld(0, pix2);
        temp_hv = vec_ld(16, pix2);
        pix2v = vec_perm(temp_lv, temp_hv, perm2vA);
        pix2 += i_stride;

        temp_lv = vec_ld(0, pix3);
        temp_hv = vec_ld(16, pix3);
        pix3v = vec_perm(temp_lv, temp_hv, perm3vA);
        pix3 += i_stride;

        sum0v = (vec_s32_t) vec_sum4s( vec_sub( vec_max( fencv, pix0v ), vec_min( fencv, pix0v ) ), (vec_u32_t) sum0v );

        sum1v = (vec_s32_t) vec_sum4s( vec_sub( vec_max( fencv, pix1v ), vec_min( fencv, pix1v ) ), (vec_u32_t) sum1v );

        sum2v = (vec_s32_t) vec_sum4s( vec_sub( vec_max( fencv, pix2v ), vec_min( fencv, pix2v ) ), (vec_u32_t) sum2v );

        sum3v = (vec_s32_t) vec_sum4s( vec_sub( vec_max( fencv, pix3v ), vec_min( fencv, pix3v ) ), (vec_u32_t) sum3v );

        temp_lv = vec_ld(0, pix0);
        temp_hv = vec_ld(16, pix0);
        pix0v = vec_perm(temp_lv, temp_hv, perm0vB);
        pix0 += i_stride;

        temp_lv = vec_ld(0, pix1);
        temp_hv = vec_ld(16, pix1);
        pix1v = vec_perm(temp_lv, temp_hv, perm1vB);
        pix1 += i_stride;

        fencv = vec_ld(0, fenc);
        fenc += FENC_STRIDE;

        temp_lv = vec_ld(0, pix2);
        temp_hv = vec_ld(16, pix2);
        pix2v = vec_perm(temp_lv, temp_hv, perm2vB);
        pix2 += i_stride;

        temp_lv = vec_ld(0, pix3);
        temp_hv = vec_ld(16, pix3);
        pix3v = vec_perm(temp_lv, temp_hv, perm3vB);
        pix3 += i_stride;

        sum0v = (vec_s32_t) vec_sum4s( vec_sub( vec_max( fencv, pix0v ), vec_min( fencv, pix0v ) ), (vec_u32_t) sum0v );

        sum1v = (vec_s32_t) vec_sum4s( vec_sub( vec_max( fencv, pix1v ), vec_min( fencv, pix1v ) ), (vec_u32_t) sum1v );

        sum2v = (vec_s32_t) vec_sum4s( vec_sub( vec_max( fencv, pix2v ), vec_min( fencv, pix2v ) ), (vec_u32_t) sum2v );

        sum3v = (vec_s32_t) vec_sum4s( vec_sub( vec_max( fencv, pix3v ), vec_min( fencv, pix3v ) ), (vec_u32_t) sum3v );