summaryrefslogtreecommitdiffstats
path: root/third_party/aom/av1/common/x86/highbd_txfm_utility_sse4.h
blob: 6f24e5948af5235d072156fc5a06f81cec39b59d (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
/*
 * Copyright (c) 2016, Alliance for Open Media. All rights reserved
 *
 * This source code is subject to the terms of the BSD 2 Clause License and
 * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
 * was not distributed with this source code in the LICENSE file, you can
 * obtain it at www.aomedia.org/license/software. If the Alliance for Open
 * Media Patent License 1.0 was not distributed with this source code in the
 * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
 */

#ifndef AOM_AV1_COMMON_X86_HIGHBD_TXFM_UTILITY_SSE4_H_
#define AOM_AV1_COMMON_X86_HIGHBD_TXFM_UTILITY_SSE4_H_

#include <smmintrin.h> /* SSE4.1 */

#define TRANSPOSE_4X4(x0, x1, x2, x3, y0, y1, y2, y3) \
  do {                                                \
    __m128i u0, u1, u2, u3;                           \
    u0 = _mm_unpacklo_epi32(x0, x1);                  \
    u1 = _mm_unpackhi_epi32(x0, x1);                  \
    u2 = _mm_unpacklo_epi32(x2, x3);                  \
    u3 = _mm_unpackhi_epi32(x2, x3);                  \
    y0 = _mm_unpacklo_epi64(u0, u2);                  \
    y1 = _mm_unpackhi_epi64(u0, u2);                  \
    y2 = _mm_unpacklo_epi64(u1, u3);                  \
    y3 = _mm_unpackhi_epi64(u1, u3);                  \
  } while (0)

static INLINE void transpose_8x8(const __m128i *in, __m128i *out) {
  TRANSPOSE_4X4(in[0], in[2], in[4], in[6], out[0], out[2], out[4], out[6]);
  TRANSPOSE_4X4(in[1], in[3], in[5], in[7], out[8], out[10], out[12], out[14]);
  TRANSPOSE_4X4(in[8], in[10], in[12], in[14], out[1], out[3], out[5], out[7]);
  TRANSPOSE_4X4(in[9], in[11], in[13], in[15], out[9], out[11], out[13],
                out[15]);
}

static INLINE void transpose_16x16(const __m128i *in, __m128i *out) {
  // Upper left 8x8
  TRANSPOSE_4X4(in[0], in[4], in[8], in[12], out[0], out[4], out[8], out[12]);
  TRANSPOSE_4X4(in[1], in[5], in[9], in[13], out[16], out[20], out[24],
                out[28]);
  TRANSPOSE_4X4(in[16], in[20], in[24], in[28], out[1], out[5], out[9],
                out[13]);
  TRANSPOSE_4X4(in[17], in[21], in[25], in[29], out[17], out[21], out[25],
                out[29]);

  // Upper right 8x8
  TRANSPOSE_4X4(in[2], in[6], in[10], in[14], out[32], out[36], out[40],
                out[44]);
  TRANSPOSE_4X4(in[3], in[7], in[11], in[15], out[48], out[52], out[56],
                out[60]);
  TRANSPOSE_4X4(in[18], in[22], in[26], in[30], out[33], out[37], out[41],
                out[45]);
  TRANSPOSE_4X4(in[19], in[23], in[27], in[31], out[49], out[53], out[57],
                out[61]);

  // Lower left 8x8
  TRANSPOSE_4X4(in[32], in[36], in[40], in[44], out[2], out[6], out[10],
                out[14]);
  TRANSPOSE_4X4(in[33], in[37], in[41], in[45], out[18], out[22], out[26],
                out[30]);
  TRANSPOSE_4X4(in[48], in[52], in[56], in[60], out[3], out[7], out[11],
                out[15]);
  TRANSPOSE_4X4(in[49], in[53], in[57], in[61], out[19], out[23], out[27],
                out[31]);
  // Lower right 8x8
  TRANSPOSE_4X4(in[34], in[38], in[42], in[46], out[34], out[38], out[42],
                out[46]);
  TRANSPOSE_4X4(in[35], in[39], in[43], in[47], out[50], out[54], out[58],
                out[62]);
  TRANSPOSE_4X4(in[50], in[54], in[58], in[62], out[35], out[39], out[43],
                out[47]);
  TRANSPOSE_4X4(in[51], in[55], in[59], in[63], out[51], out[55], out[59],
                out[63]);
}

static INLINE void transpose_32x32(const __m128i *input, __m128i *output) {
  for (int j = 0; j < 8; j++) {
    for (int i = 0; i < 8; i++) {
      TRANSPOSE_4X4(input[i * 32 + j + 0], input[i * 32 + j + 8],
                    input[i * 32 + j + 16], input[i * 32 + j + 24],
                    output[j * 32 + i + 0], output[j * 32 + i + 8],
                    output[j * 32 + i + 16], output[j * 32 + i + 24]);
    }
  }
}

// Note:
//  rounding = 1 << (bit - 1)
static INLINE __m128i half_btf_sse4_1(const __m128i *w0, const __m128i *n0,
                                      const __m128i *w1, const __m128i *n1,
                                      const __m128i *rounding, int bit) {
  __m128i x, y;

  x = _mm_mullo_epi32(*w0, *n0);
  y = _mm_mullo_epi32(*w1, *n1);
  x = _mm_add_epi32(x, y);
  x = _mm_add_epi32(x, *rounding);
  x = _mm_srai_epi32(x, bit);
  return x;
}

static INLINE __m128i half_btf_0_sse4_1(const __m128i *w0, const __m128i *n0,
                                        const __m128i *rounding, int bit) {
  __m128i x;

  x = _mm_mullo_epi32(*w0, *n0);
  x = _mm_add_epi32(x, *rounding);
  x = _mm_srai_epi32(x, bit);
  return x;
}

typedef void (*transform_1d_sse4_1)(__m128i *in, __m128i *out, int bit,
                                    int do_cols, int bd, int out_shift);

typedef void (*fwd_transform_1d_sse4_1)(__m128i *in, __m128i *out, int bit,
                                        const int num_cols);

void av1_highbd_inv_txfm2d_add_universe_sse4_1(const int32_t *input,
                                               uint8_t *output, int stride,
                                               TX_TYPE tx_type, TX_SIZE tx_size,
                                               int eob, const int bd);

#endif  // AOM_AV1_COMMON_X86_HIGHBD_TXFM_UTILITY_SSE4_H_