1 /*
2 * Copyright (c) 2016, Alliance for Open Media. All rights reserved
3 *
4 * This source code is subject to the terms of the BSD 2 Clause License and
5 * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
6 * was not distributed with this source code in the LICENSE file, you can
7 * obtain it at www.aomedia.org/license/software. If the Alliance for Open
8 * Media Patent License 1.0 was not distributed with this source code in the
9 * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
10 */
11
12 #include <arm_neon.h>
13
14 #include "aom_dsp/inv_txfm.h"
15 #include "aom_ports/mem.h"
16
aom_idct8x8_1_add_neon(int16_t * input,uint8_t * dest,int dest_stride)17 void aom_idct8x8_1_add_neon(int16_t *input, uint8_t *dest, int dest_stride) {
18 uint8x8_t d2u8, d3u8, d30u8, d31u8;
19 uint64x1_t d2u64, d3u64, d4u64, d5u64;
20 uint16x8_t q0u16, q9u16, q10u16, q11u16, q12u16;
21 int16x8_t q0s16;
22 uint8_t *d1, *d2;
23 int16_t i, a1;
24 int16_t out = dct_const_round_shift(input[0] * cospi_16_64);
25 out = dct_const_round_shift(out * cospi_16_64);
26 a1 = ROUND_POWER_OF_TWO(out, 5);
27
28 q0s16 = vdupq_n_s16(a1);
29 q0u16 = vreinterpretq_u16_s16(q0s16);
30
31 d1 = d2 = dest;
32 for (i = 0; i < 2; i++) {
33 d2u64 = vld1_u64((const uint64_t *)d1);
34 d1 += dest_stride;
35 d3u64 = vld1_u64((const uint64_t *)d1);
36 d1 += dest_stride;
37 d4u64 = vld1_u64((const uint64_t *)d1);
38 d1 += dest_stride;
39 d5u64 = vld1_u64((const uint64_t *)d1);
40 d1 += dest_stride;
41
42 q9u16 = vaddw_u8(q0u16, vreinterpret_u8_u64(d2u64));
43 q10u16 = vaddw_u8(q0u16, vreinterpret_u8_u64(d3u64));
44 q11u16 = vaddw_u8(q0u16, vreinterpret_u8_u64(d4u64));
45 q12u16 = vaddw_u8(q0u16, vreinterpret_u8_u64(d5u64));
46
47 d2u8 = vqmovun_s16(vreinterpretq_s16_u16(q9u16));
48 d3u8 = vqmovun_s16(vreinterpretq_s16_u16(q10u16));
49 d30u8 = vqmovun_s16(vreinterpretq_s16_u16(q11u16));
50 d31u8 = vqmovun_s16(vreinterpretq_s16_u16(q12u16));
51
52 vst1_u64((uint64_t *)d2, vreinterpret_u64_u8(d2u8));
53 d2 += dest_stride;
54 vst1_u64((uint64_t *)d2, vreinterpret_u64_u8(d3u8));
55 d2 += dest_stride;
56 vst1_u64((uint64_t *)d2, vreinterpret_u64_u8(d30u8));
57 d2 += dest_stride;
58 vst1_u64((uint64_t *)d2, vreinterpret_u64_u8(d31u8));
59 d2 += dest_stride;
60 }
61 return;
62 }
63