void test_vst1u32 (void) { uint32_t *arg0_uint32_t; uint32x2_t arg1_uint32x2_t; vst1_u32 (arg0_uint32_t, arg1_uint32x2_t); }
// 'do_above' and 'do_left' facilitate branch removal when inlined. static INLINE void dc_8x8(uint8_t *dst, ptrdiff_t stride, const uint8_t *above, const uint8_t *left, int do_above, int do_left) { uint16x8_t sum_top; uint16x8_t sum_left; uint8x8_t dc0; if (do_above) { const uint8x8_t A = vld1_u8(above); // top row const uint16x4_t p0 = vpaddl_u8(A); // cascading summation of the top const uint16x4_t p1 = vpadd_u16(p0, p0); const uint16x4_t p2 = vpadd_u16(p1, p1); sum_top = vcombine_u16(p2, p2); } if (do_left) { const uint8x8_t L = vld1_u8(left); // left border const uint16x4_t p0 = vpaddl_u8(L); // cascading summation of the left const uint16x4_t p1 = vpadd_u16(p0, p0); const uint16x4_t p2 = vpadd_u16(p1, p1); sum_left = vcombine_u16(p2, p2); } if (do_above && do_left) { const uint16x8_t sum = vaddq_u16(sum_left, sum_top); dc0 = vrshrn_n_u16(sum, 4); } else if (do_above) { dc0 = vrshrn_n_u16(sum_top, 3); } else if (do_left) { dc0 = vrshrn_n_u16(sum_left, 3); } else { dc0 = vdup_n_u8(0x80); } { const uint8x8_t dc = vdup_lane_u8(dc0, 0); int i; for (i = 0; i < 8; ++i) { vst1_u32((uint32_t*)(dst + i * stride), vreinterpret_u32_u8(dc)); } } }
inline void vst1(u32 * ptr, const uint32x2_t & v) { return vst1_u32(ptr, v); }