void vpx_fdct16x16_msa(const int16_t *input, int16_t *output,
int32_t src_stride) {
int32_t i;
DECLARE_ALIGNED(32, int16_t, tmp_buf[16 * 16]);
/* column transform */
for (i = 0; i < 2; ++i) {
fdct8x16_1d_column((input + 8 * i), (&tmp_buf[0] + 8 * i), src_stride);
}
/* row transform */
for (i = 0; i < 2; ++i) {
fdct16x8_1d_row((&tmp_buf[0] + (128 * i)), (output + (128 * i)));
}
}
void av1_fht16x16_msa(const int16_t *input, int16_t *output, int32_t stride,
int32_t tx_type) {
DECLARE_ALIGNED(32, int16_t, tmp[256]);
DECLARE_ALIGNED(32, int16_t, trans_buf[256]);
DECLARE_ALIGNED(32, int16_t, tmp_buf[128]);
int32_t i;
int16_t *ptmpbuf = &tmp_buf[0];
int16_t *trans = &trans_buf[0];
const int32_t const_arr[29 * 4] = {
52707308, 52707308, 52707308, 52707308, -1072430300,
-1072430300, -1072430300, -1072430300, 795618043, 795618043,
795618043, 795618043, -721080468, -721080468, -721080468,
-721080468, 459094491, 459094491, 459094491, 459094491,
-970646691, -970646691, -970646691, -970646691, 1010963856,
1010963856, 1010963856, 1010963856, -361743294, -361743294,
-361743294, -361743294, 209469125, 209469125, 209469125,
209469125, -1053094788, -1053094788, -1053094788, -1053094788,
1053160324, 1053160324, 1053160324, 1053160324, 639644520,
639644520, 639644520, 639644520, -862444000, -862444000,
-862444000, -862444000, 1062144356, 1062144356, 1062144356,
1062144356, -157532337, -157532337, -157532337, -157532337,
260914709, 260914709, 260914709, 260914709, -1041559667,
-1041559667, -1041559667, -1041559667, 920985831, 920985831,
920985831, 920985831, -551995675, -551995675, -551995675,
-551995675, 596522295, 596522295, 596522295, 596522295,
892853362, 892853362, 892853362, 892853362, -892787826,
-892787826, -892787826, -892787826, 410925857, 410925857,
410925857, 410925857, -992012162, -992012162, -992012162,
-992012162, 992077698, 992077698, 992077698, 992077698,
759246145, 759246145, 759246145, 759246145, -759180609,
-759180609, -759180609, -759180609, -759222975, -759222975,
-759222975, -759222975, 759288511, 759288511, 759288511,
759288511
};
switch (tx_type) {
case DCT_DCT:
/* column transform */
for (i = 0; i < 2; ++i) {
fdct8x16_1d_column(input + 8 * i, tmp + 8 * i, stride);
}
/* row transform */
for (i = 0; i < 2; ++i) {
fdct16x8_1d_row(tmp + (128 * i), output + (128 * i));
}
break;
case ADST_DCT:
/* column transform */
for (i = 0; i < 2; ++i) {
fadst16_cols_step1_msa(input + (i << 3), stride, const_arr, ptmpbuf);
fadst16_cols_step2_msa(ptmpbuf, const_arr, tmp + (i << 3));
}
/* row transform */
for (i = 0; i < 2; ++i) {
postproc_fdct16x8_1d_row(tmp + (128 * i), output + (128 * i));
}
break;
case DCT_ADST:
/* column transform */
for (i = 0; i < 2; ++i) {
fdct8x16_1d_column(input + 8 * i, tmp + 8 * i, stride);
}
fadst16_transpose_postproc_msa(tmp, trans);
/* row transform */
for (i = 0; i < 2; ++i) {
fadst16_rows_step1_msa(trans + (i << 7), const_arr, ptmpbuf);
fadst16_rows_step2_msa(ptmpbuf, const_arr, tmp + (i << 7));
}
fadst16_transpose_msa(tmp, output);
break;
case ADST_ADST:
/* column transform */
for (i = 0; i < 2; ++i) {
fadst16_cols_step1_msa(input + (i << 3), stride, const_arr, ptmpbuf);
fadst16_cols_step2_msa(ptmpbuf, const_arr, tmp + (i << 3));
}
fadst16_transpose_postproc_msa(tmp, trans);
/* row transform */
for (i = 0; i < 2; ++i) {
fadst16_rows_step1_msa(trans + (i << 7), const_arr, ptmpbuf);
fadst16_rows_step2_msa(ptmpbuf, const_arr, tmp + (i << 7));
}
fadst16_transpose_msa(tmp, output);
break;
default: assert(0); break;
}
}