C++ (Cpp) HADD_UH_U32 Examples

Example #1

File: enc_msa.c Project: garrettmoon/libwebp

static WEBP_INLINE void DCMode16x16(uint8_t* dst, const uint8_t* left,
                                    const uint8_t* top) {
  int DC;
  v16u8 out;
  if (top != NULL && left != NULL) {
    const v16u8 rtop = LD_UB(top);
    const v8u16 dctop = __msa_hadd_u_h(rtop, rtop);
    const v16u8 rleft = LD_UB(left);
    const v8u16 dcleft = __msa_hadd_u_h(rleft, rleft);
    const v8u16 dctemp = dctop + dcleft;
    DC = HADD_UH_U32(dctemp);
    DC = (DC + 16) >> 5;
  } else if (left != NULL) {   // left but no top

Example #2

File: dct-c.c Project: 0x0B501E7E/x264

static int32_t subtract_sum4x4_msa( uint8_t *p_src, int32_t i_src_stride,
                                    uint8_t *pred_ptr, int32_t i_pred_stride )
{
    int16_t i_sum;
    uint32_t i_src0, i_src1, i_src2, i_src3;
    uint32_t i_pred0, i_pred1, i_pred2, i_pred3;
    v16i8 src = { 0 };
    v16i8 pred = { 0 };
    v16u8 src_l0, src_l1;
    v8i16 diff0, diff1;

    LW4( p_src, i_src_stride, i_src0, i_src1, i_src2, i_src3 );
    LW4( pred_ptr, i_pred_stride, i_pred0, i_pred1, i_pred2, i_pred3 );
    INSERT_W4_SB( i_src0, i_src1, i_src2, i_src3, src );
    INSERT_W4_SB( i_pred0, i_pred1, i_pred2, i_pred3, pred );
    ILVRL_B2_UB( src, pred, src_l0, src_l1 );
    HSUB_UB2_SH( src_l0, src_l1, diff0, diff1 );
    i_sum = HADD_UH_U32( diff0 + diff1 );

    return i_sum;
}

Example #3

File: mpegvideoencdsp_msa.c Project: 0day-ci/FFmpeg

static int32_t sum_u8src_16width_msa(uint8_t *src, int32_t stride)
{
    uint32_t sum = 0;
    v16u8 in0, in1, in2, in3, in4, in5, in6, in7;
    v16u8 in8, in9, in10, in11, in12, in13, in14, in15;

    LD_UB8(src, stride, in0, in1, in2, in3, in4, in5, in6, in7);
    src += (8 * stride);
    LD_UB8(src, stride, in8, in9, in10, in11, in12, in13, in14, in15);

    HADD_UB4_UB(in0, in1, in2, in3, in0, in1, in2, in3);
    HADD_UB4_UB(in4, in5, in6, in7, in4, in5, in6, in7);
    HADD_UB4_UB(in8, in9, in10, in11, in8, in9, in10, in11);
    HADD_UB4_UB(in12, in13, in14, in15, in12, in13, in14, in15);

    sum = HADD_UH_U32(in0);
    sum += HADD_UH_U32(in1);
    sum += HADD_UH_U32(in2);
    sum += HADD_UH_U32(in3);
    sum += HADD_UH_U32(in4);
    sum += HADD_UH_U32(in5);
    sum += HADD_UH_U32(in6);
    sum += HADD_UH_U32(in7);
    sum += HADD_UH_U32(in8);
    sum += HADD_UH_U32(in9);
    sum += HADD_UH_U32(in10);
    sum += HADD_UH_U32(in11);
    sum += HADD_UH_U32(in12);
    sum += HADD_UH_U32(in13);
    sum += HADD_UH_U32(in14);
    sum += HADD_UH_U32(in15);

    return sum;
}

Example #4

File: enc_msa.c Project: garrettmoon/libwebp

static WEBP_INLINE void DCMode16x16(uint8_t* dst, const uint8_t* left,
                                    const uint8_t* top) {
  int DC;
  v16u8 out;
  if (top != NULL && left != NULL) {
    const v16u8 rtop = LD_UB(top);
    const v8u16 dctop = __msa_hadd_u_h(rtop, rtop);
    const v16u8 rleft = LD_UB(left);
    const v8u16 dcleft = __msa_hadd_u_h(rleft, rleft);
    const v8u16 dctemp = dctop + dcleft;
    DC = HADD_UH_U32(dctemp);
    DC = (DC + 16) >> 5;
  } else if (left != NULL) {   // left but no top
    const v16u8 rleft = LD_UB(left);
    const v8u16 dcleft = __msa_hadd_u_h(rleft, rleft);
    DC = HADD_UH_U32(dcleft);
    DC = (DC + DC + 16) >> 5;
  } else if (top != NULL) {   // top but no left
    const v16u8 rtop = LD_UB(top);
    const v8u16 dctop = __msa_hadd_u_h(rtop, rtop);
    DC = HADD_UH_U32(dctop);
    DC = (DC + DC + 16) >> 5;
  } else {   // no top, no left, nothing.
    DC = 0x80;
  }
  out = (v16u8)__msa_fill_b(DC);
  STORE16x16(out, dst);
}

static void Intra16Preds(uint8_t* dst,
                         const uint8_t* left, const uint8_t* top) {