void vpx_idct4x4_1_add_neon(const tran_low_t *input, uint8_t *dest,
int stride) {
const int16_t out0 = WRAPLOW(dct_const_round_shift(input[0] * cospi_16_64));
const int16_t out1 = WRAPLOW(dct_const_round_shift(out0 * cospi_16_64));
const int16_t a1 = ROUND_POWER_OF_TWO(out1, 4);
const int16x8_t dc = vdupq_n_s16(a1);
uint32x2_t d = vdup_n_u32(0);
assert(!((intptr_t)dest % sizeof(uint32_t)));
assert(!(stride % sizeof(uint32_t)));
idct4x4_1_add_kernel(&dest, stride, dc, &d);
idct4x4_1_add_kernel(&dest, stride, dc, &d);
}
void vpx_idct32x32_1_add_neon(const tran_low_t *input, uint8_t *dest,
int stride) {
int i;
const int16_t out0 =
WRAPLOW(dct_const_round_shift((int16_t)input[0] * cospi_16_64));
const int16_t out1 = WRAPLOW(dct_const_round_shift(out0 * cospi_16_64));
const int16_t a1 = ROUND_POWER_OF_TWO(out1, 6);
if (a1 >= 0) {
const uint8x16_t dc = create_dcq(a1);
for (i = 0; i < 32; i++) {
idct32x32_1_add_pos_kernel(&dest, stride, dc);
}
} else {
const uint8x16_t dc = create_dcq(-a1);
for (i = 0; i < 32; i++) {
idct32x32_1_add_neg_kernel(&dest, stride, dc);
}
}
}
