void x264_sub4x4_dct_altivec( int16_t dct[4][4],
uint8_t *pix1, int i_pix1, uint8_t *pix2, int i_pix2 )
{
PREP_DIFF;
PREP_STORE8;
vec_s16_t dct0v, dct1v, dct2v, dct3v;
vec_s16_t tmp0v, tmp1v, tmp2v, tmp3v;
VEC_DIFF_H( pix1, i_pix1, pix2, i_pix2, 4, dct0v );
VEC_DIFF_H( pix1, i_pix1, pix2, i_pix2, 4, dct1v );
VEC_DIFF_H( pix1, i_pix1, pix2, i_pix2, 4, dct2v );
VEC_DIFF_H( pix1, i_pix1, pix2, i_pix2, 4, dct3v );
VEC_DCT( dct0v, dct1v, dct2v, dct3v, tmp0v, tmp1v, tmp2v, tmp3v );
VEC_TRANSPOSE_4( tmp0v, tmp1v, tmp2v, tmp3v,
dct0v, dct1v, dct2v, dct3v );
VEC_DCT( dct0v, dct1v, dct2v, dct3v, tmp0v, tmp1v, tmp2v, tmp3v );
VEC_TRANSPOSE_4( tmp0v, tmp1v, tmp2v, tmp3v,
dct0v, dct1v, dct2v, dct3v );
VEC_STORE8( dct0v, dct[0] );
VEC_STORE8( dct1v, dct[1] );
VEC_STORE8( dct2v, dct[2] );
VEC_STORE8( dct3v, dct[3] );
}
static void predict_8x8c_p_altivec( uint8_t *src )
{
int H = 0, V = 0;
for( int i = 0; i < 4; i++ )
{
H += ( i + 1 ) * ( src[4+i - FDEC_STRIDE] - src[2 - i -FDEC_STRIDE] );
V += ( i + 1 ) * ( src[-1 +(i+4)*FDEC_STRIDE] - src[-1+(2-i)*FDEC_STRIDE] );
}
int a = 16 * ( src[-1+7*FDEC_STRIDE] + src[7 - FDEC_STRIDE] );
int b = ( 17 * H + 16 ) >> 5;
int c = ( 17 * V + 16 ) >> 5;
int i00 = a -3*b -3*c + 16;
vec_s16_u i00_u, b_u, c_u;
i00_u.s[0] = i00;
b_u.s[0] = b;
c_u.s[0] = c;
vec_u16_t val5_v = vec_splat_u16(5);
vec_s16_t i00_v, b_v, c_v;
i00_v = vec_splat(i00_u.v, 0);
b_v = vec_splat(b_u.v, 0);
c_v = vec_splat(c_u.v, 0);
vec_s16_t induc_v = (vec_s16_t) CV(0, 1, 2, 3, 4, 5, 6, 7);
vec_s16_t add_i0_b_0v = vec_mladd(induc_v, b_v, i00_v);
PREP_STORE8;
for( int i = 0; i < 8; ++i )
{
vec_s16_t shift_0_v = vec_sra(add_i0_b_0v, val5_v);
vec_u8_t com_sat_v = vec_packsu(shift_0_v, shift_0_v);
VEC_STORE8(com_sat_v, &src[0]);
src += FDEC_STRIDE;
add_i0_b_0v = vec_adds(add_i0_b_0v, c_v);
}
}