uint16x8_t g0 = vaddq_u16(vshrq_n_u16(r0, 8), vshrq_n_u16(r2, 8));
uint16x8_t g1 = vandq_u16(r1, masklo);
g0 = vaddq_u16(g0, vaddq_u16(g1, vextq_u16(g1, g1, 1)));
- uint16x8_t rot = vextq_u16(g1, g1, 1)
+ uint16x8_t rot = vextq_u16(g1, g1, 1);
g1 = vshlq_n_u16(rot, 2);
// g0 = b0 b2 b4 ...
// g1 = b1 b3 b5 ...
v_r01 = vaddq_s32(vqaddq_s32(v_r01, v_r41), vqaddq_s32(v_r21, v_r21));
v_r11 = vaddq_s32(vqaddq_s32(v_r11, v_r21), v_r31);
- shifted = vshlq_n_s32(v_r11, 2)
+ shifted = vshlq_n_s32(v_r11, 2);
int32x4_t v_dst1 = vshrq_n_s32(vaddq_s32(vqaddq_s32(v_r01, shifted), v_delta), 8);
vst1q_u16(dst + x, vcombine_u16(vqmovun_s32(v_dst0), vqmovun_s32(v_dst1)));