static av_always_inline void h264_idct_dc_add_internal(uint8_t *dst, DCTELEM *block, int stride, int size)
{
vec_s16 dc16;
vec_u8 dcplus, dcminus, v0, v1, v2, v3, aligner;
LOAD_ZERO;
DECLARE_ALIGNED(16, int, dc);
int i;
dc = (block[0] + 32) >> 6;
dc16 = __vsplth(__lvewx(&dc, 0), 1);
if (size == 4)
dc16 = __vsldoi(dc16, zero_s16v, 8);
dcplus = __vpkshus(dc16, zero_s16v);
dcminus = __vpkshus(__vsubuhm(zero_s16v, dc16), zero_s16v);
aligner = __lvsr(0, dst);
dcplus = __perm(dcplus, dcplus, aligner);
dcminus = __perm(dcminus, dcminus, aligner);
for (i = 0; i < size; i += 4) {
v0 = __lvx(dst+0*stride, 0);
v1 = __lvx(dst+1*stride, 0);
v2 = __lvx(dst+2*stride, 0);
v3 = __lvx(dst+3*stride, 0);
v0 = __vaddubs(v0, dcplus);
v1 = __vaddubs(v1, dcplus);
v2 = __vaddubs(v2, dcplus);
v3 = __vaddubs(v3, dcplus);
v0 = __vsububs(v0, dcminus);
v1 = __vsububs(v1, dcminus);
v2 = __vsububs(v2, dcminus);
v3 = __vsububs(v3, dcminus);
__stvx(v0, dst+0*stride, 0);
__stvx(v1, dst+1*stride, 0);
__stvx(v2, dst+2*stride, 0);
__stvx(v3, dst+3*stride, 0);
dst += 4*stride;
}
}
/* dest*~srca + src */
static force_inline __vector4
over (__vector4 src,
__vector4 srca,
__vector4 dest)
{
__vector4 tmp = pix_multiply (dest, negate (srca));
tmp = __vaddubs (src, tmp);
return tmp;
}
static force_inline __vector4
pix_add (__vector4 a, __vector4 b)
{
return __vaddubs(a,b);
}
