void aom_iwht4x4_1_add_msa(const int16_t *input, uint8_t *dst,
int32_t dst_stride) {
int16_t a1, e1;
v8i16 in1, in0 = { 0 };
a1 = input[0] >> UNIT_QUANT_SHIFT;
e1 = a1 >> 1;
a1 -= e1;
in0 = __msa_insert_h(in0, 0, a1);
in0 = __msa_insert_h(in0, 1, e1);
in0 = __msa_insert_h(in0, 2, e1);
in0 = __msa_insert_h(in0, 3, e1);
in1 = in0 >> 1;
in0 -= in1;
ADDBLK_ST4x4_UB(in0, in1, in1, in1, dst, dst_stride);
}
static void FTransformWHT(const int16_t* in, int16_t* out) {
v8i16 in0 = { 0 };
v8i16 in1 = { 0 };
v8i16 tmp0, tmp1, tmp2, tmp3;
v8i16 out0, out1;
const v8i16 mask0 = { 0, 1, 2, 3, 8, 9, 10, 11 };
const v8i16 mask1 = { 4, 5, 6, 7, 12, 13, 14, 15 };
const v8i16 mask2 = { 0, 4, 8, 12, 1, 5, 9, 13 };
const v8i16 mask3 = { 3, 7, 11, 15, 2, 6, 10, 14 };
in0 = __msa_insert_h(in0, 0, in[ 0]);
in0 = __msa_insert_h(in0, 1, in[ 64]);
in0 = __msa_insert_h(in0, 2, in[128]);
in0 = __msa_insert_h(in0, 3, in[192]);
in0 = __msa_insert_h(in0, 4, in[ 16]);
in0 = __msa_insert_h(in0, 5, in[ 80]);
in0 = __msa_insert_h(in0, 6, in[144]);
in0 = __msa_insert_h(in0, 7, in[208]);
in1 = __msa_insert_h(in1, 0, in[ 48]);
in1 = __msa_insert_h(in1, 1, in[112]);
in1 = __msa_insert_h(in1, 2, in[176]);
in1 = __msa_insert_h(in1, 3, in[240]);
in1 = __msa_insert_h(in1, 4, in[ 32]);
in1 = __msa_insert_h(in1, 5, in[ 96]);
in1 = __msa_insert_h(in1, 6, in[160]);
in1 = __msa_insert_h(in1, 7, in[224]);
ADDSUB2(in0, in1, tmp0, tmp1);
VSHF_H2_SH(tmp0, tmp1, tmp0, tmp1, mask0, mask1, tmp2, tmp3);
ADDSUB2(tmp2, tmp3, tmp0, tmp1);
VSHF_H2_SH(tmp0, tmp1, tmp0, tmp1, mask2, mask3, in0, in1);
ADDSUB2(in0, in1, tmp0, tmp1);
VSHF_H2_SH(tmp0, tmp1, tmp0, tmp1, mask0, mask1, tmp2, tmp3);
ADDSUB2(tmp2, tmp3, out0, out1);
SRAI_H2_SH(out0, out1, 1);
ST_SH2(out0, out1, out, 8);
}