vec_s8 filter_outerl = vec_sld(filter_outerh, filter_outerh, 2)
#define FILTER_H(dstv, off) \
- a = vec_ld((off)-2, src); \
- b = vec_ld((off)-2+15, src); \
+ a = vec_ld((off)-is6tap-1, src); \
+ b = vec_ld((off)-is6tap-1+15, src); \
\
pixh = vec_perm(a, b, permh##off); \
pixl = vec_perm(a, b, perml##off); \
vec_s16 f16h, f16l;
vec_s32 filth, filtl;
- vec_u8 perm_inner = { 1,2,3,4, 2,3,4,5, 3,4,5,6, 4,5,6,7 };
+ vec_u8 perm_inner6 = { 1,2,3,4, 2,3,4,5, 3,4,5,6, 4,5,6,7 };
+ vec_u8 perm_inner4 = { 0,1,2,3, 1,2,3,4, 2,3,4,5, 3,4,5,6 };
+ vec_u8 perm_inner = is6tap ? perm_inner6 : perm_inner4;
vec_u8 perm_outer = { 4,9, 0,5, 5,10, 1,6, 6,11, 2,7, 7,12, 3,8 };
vec_s32 c64 = vec_sl(vec_splat_s32(1), vec_splat_u32(6));
vec_u16 c7 = vec_splat_u16(7);
- align_vec0 = vec_lvsl( -2, src);
- align_vec8 = vec_lvsl(8-2, src);
+ align_vec0 = vec_lvsl( -is6tap-1, src);
+ align_vec8 = vec_lvsl(8-is6tap-1, src);
permh0 = vec_perm(align_vec0, align_vec0, perm_inner);
permh8 = vec_perm(align_vec8, align_vec8, perm_inner);
static void put_vp8_epel ## WIDTH ## _h ## HTAPS ## v ## VTAPS ## _altivec(uint8_t *dst, int stride, uint8_t *src, int s, int h, int mx, int my) \
{ \
DECLARE_ALIGNED(16, uint8_t, tmp)[(2*WIDTH+5)*16]; \
- put_vp8_epel ## WIDTH ## _h ## HTAPS ## _altivec(tmp, 16, src-2*stride, stride, h+5, mx, my); \
- put_vp8_epel ## WIDTH ## _v ## VTAPS ## _altivec(dst, stride, tmp+2*16, 16, h, mx, my); \
+ if (VTAPS == 6) { \
+ put_vp8_epel ## WIDTH ## _h ## HTAPS ## _altivec(tmp, 16, src-2*stride, stride, h+5, mx, my); \
+ put_vp8_epel ## WIDTH ## _v ## VTAPS ## _altivec(dst, stride, tmp+2*16, 16, h, mx, my); \
+ } else { \
+ put_vp8_epel ## WIDTH ## _h ## HTAPS ## _altivec(tmp, 16, src-stride, stride, h+4, mx, my); \
+ put_vp8_epel ## WIDTH ## _v ## VTAPS ## _altivec(dst, stride, tmp+16, 16, h, mx, my); \
+ } \
}
EPEL_FUNCS(16,6)