1/*
2 *  Copyright (c) 2014 The WebM project authors. All Rights Reserved.
3 *
4 *  Use of this source code is governed by a BSD-style license
5 *  that can be found in the LICENSE file in the root of the source
6 *  tree. An additional intellectual property rights grant can be found
7 *  in the file PATENTS.  All contributing project authors may
8 *  be found in the AUTHORS file in the root of the source tree.
9 */
10
11#include <arm_neon.h>
12
13void idct_dequant_0_2x_neon(int16_t *q, int16_t dq, unsigned char *dst,
14                            int stride) {
15  unsigned char *dst0;
16  int i, a0, a1;
17  int16x8x2_t q2Add;
18  int32x2_t d2s32 = vdup_n_s32(0), d4s32 = vdup_n_s32(0);
19  uint8x8_t d2u8, d4u8;
20  uint16x8_t q1u16, q2u16;
21
22  a0 = ((q[0] * dq) + 4) >> 3;
23  a1 = ((q[16] * dq) + 4) >> 3;
24  q[0] = q[16] = 0;
25  q2Add.val[0] = vdupq_n_s16((int16_t)a0);
26  q2Add.val[1] = vdupq_n_s16((int16_t)a1);
27
28  for (i = 0; i < 2; i++, dst += 4) {
29    dst0 = dst;
30    d2s32 = vld1_lane_s32((const int32_t *)dst0, d2s32, 0);
31    dst0 += stride;
32    d2s32 = vld1_lane_s32((const int32_t *)dst0, d2s32, 1);
33    dst0 += stride;
34    d4s32 = vld1_lane_s32((const int32_t *)dst0, d4s32, 0);
35    dst0 += stride;
36    d4s32 = vld1_lane_s32((const int32_t *)dst0, d4s32, 1);
37
38    q1u16 = vaddw_u8(vreinterpretq_u16_s16(q2Add.val[i]),
39                     vreinterpret_u8_s32(d2s32));
40    q2u16 = vaddw_u8(vreinterpretq_u16_s16(q2Add.val[i]),
41                     vreinterpret_u8_s32(d4s32));
42
43    d2u8 = vqmovun_s16(vreinterpretq_s16_u16(q1u16));
44    d4u8 = vqmovun_s16(vreinterpretq_s16_u16(q2u16));
45
46    d2s32 = vreinterpret_s32_u8(d2u8);
47    d4s32 = vreinterpret_s32_u8(d4u8);
48
49    dst0 = dst;
50    vst1_lane_s32((int32_t *)dst0, d2s32, 0);
51    dst0 += stride;
52    vst1_lane_s32((int32_t *)dst0, d2s32, 1);
53    dst0 += stride;
54    vst1_lane_s32((int32_t *)dst0, d4s32, 0);
55    dst0 += stride;
56    vst1_lane_s32((int32_t *)dst0, d4s32, 1);
57  }
58  return;
59}
60