summaryrefslogtreecommitdiffstats
path: root/third_party/aom/aom_dsp/x86/fwd_txfm_sse2.c
blob: a337e618deb984717f8a5613a6a975133909ee08 (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
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
/*
 * 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.
 */

#include <emmintrin.h>  // SSE2

#include "./aom_config.h"
#include "./aom_dsp_rtcd.h"
#include "aom_dsp/aom_dsp_common.h"
#include "aom_dsp/x86/fwd_txfm_sse2.h"

void aom_fdct4x4_1_sse2(const int16_t *input, tran_low_t *output, int stride) {
  __m128i in0, in1;
  __m128i tmp;
  const __m128i zero = _mm_setzero_si128();
  in0 = _mm_loadl_epi64((const __m128i *)(input + 0 * stride));
  in1 = _mm_loadl_epi64((const __m128i *)(input + 1 * stride));
  in1 = _mm_unpacklo_epi64(
      in1, _mm_loadl_epi64((const __m128i *)(input + 2 * stride)));
  in0 = _mm_unpacklo_epi64(
      in0, _mm_loadl_epi64((const __m128i *)(input + 3 * stride)));

  tmp = _mm_add_epi16(in0, in1);
  in0 = _mm_unpacklo_epi16(zero, tmp);
  in1 = _mm_unpackhi_epi16(zero, tmp);
  in0 = _mm_srai_epi32(in0, 16);
  in1 = _mm_srai_epi32(in1, 16);

  tmp = _mm_add_epi32(in0, in1);
  in0 = _mm_unpacklo_epi32(tmp, zero);
  in1 = _mm_unpackhi_epi32(tmp, zero);

  tmp = _mm_add_epi32(in0, in1);
  in0 = _mm_srli_si128(tmp, 8);

  in1 = _mm_add_epi32(tmp, in0);
  in0 = _mm_slli_epi32(in1, 1);
  output[0] = (tran_low_t)_mm_cvtsi128_si32(in0);
}

void aom_fdct8x8_1_sse2(const int16_t *input, tran_low_t *output, int stride) {
  __m128i in0 = _mm_load_si128((const __m128i *)(input + 0 * stride));
  __m128i in1 = _mm_load_si128((const __m128i *)(input + 1 * stride));
  __m128i in2 = _mm_load_si128((const __m128i *)(input + 2 * stride));
  __m128i in3 = _mm_load_si128((const __m128i *)(input + 3 * stride));
  __m128i u0, u1, sum;

  u0 = _mm_add_epi16(in0, in1);
  u1 = _mm_add_epi16(in2, in3);

  in0 = _mm_load_si128((const __m128i *)(input + 4 * stride));
  in1 = _mm_load_si128((const __m128i *)(input + 5 * stride));
  in2 = _mm_load_si128((const __m128i *)(input + 6 * stride));
  in3 = _mm_load_si128((const __m128i *)(input + 7 * stride));

  sum = _mm_add_epi16(u0, u1);

  in0 = _mm_add_epi16(in0, in1);
  in2 = _mm_add_epi16(in2, in3);
  sum = _mm_add_epi16(sum, in0);

  u0 = _mm_setzero_si128();
  sum = _mm_add_epi16(sum, in2);

  in0 = _mm_unpacklo_epi16(u0, sum);
  in1 = _mm_unpackhi_epi16(u0, sum);
  in0 = _mm_srai_epi32(in0, 16);
  in1 = _mm_srai_epi32(in1, 16);

  sum = _mm_add_epi32(in0, in1);
  in0 = _mm_unpacklo_epi32(sum, u0);
  in1 = _mm_unpackhi_epi32(sum, u0);

  sum = _mm_add_epi32(in0, in1);
  in0 = _mm_srli_si128(sum, 8);

  in1 = _mm_add_epi32(sum, in0);
  output[0] = (tran_low_t)_mm_cvtsi128_si32(in1);
}

void aom_fdct16x16_1_sse2(const int16_t *input, tran_low_t *output,
                          int stride) {
  __m128i in0, in1, in2, in3;
  __m128i u0, u1;
  __m128i sum = _mm_setzero_si128();
  int i;

  for (i = 0; i < 2; ++i) {
    in0 = _mm_load_si128((const __m128i *)(input + 0 * stride + 0));
    in1 = _mm_load_si128((const __m128i *)(input + 0 * stride + 8));
    in2 = _mm_load_si128((const __m128i *)(input + 1 * stride + 0));
    in3 = _mm_load_si128((const __m128i *)(input + 1 * stride + 8));

    u0 = _mm_add_epi16(in0, in1);
    u1 = _mm_add_epi16(in2, in3);
    sum = _mm_add_epi16(sum, u0);

    in0 = _mm_load_si128((const __m128i *)(input + 2 * stride + 0));
    in1 = _mm_load_si128((const __m128i *)(input + 2 * stride + 8));
    in2 = _mm_load_si128((const __m128i *)(input + 3 * stride + 0));
    in3 = _mm_load_si128((const __m128i *)(input + 3 * stride + 8));

    sum = _mm_add_epi16(sum, u1);
    u0 = _mm_add_epi16(in0, in1);
    u1 = _mm_add_epi16(in2, in3);
    sum = _mm_add_epi16(sum, u0);

    in0 = _mm_load_si128((const __m128i *)(input + 4 * stride + 0));
    in1 = _mm_load_si128((const __m128i *)(input + 4 * stride + 8));
    in2 = _mm_load_si128((const __m128i *)(input + 5 * stride + 0));
    in3 = _mm_load_si128((const __m128i *)(input + 5 * stride + 8));

    sum = _mm_add_epi16(sum, u1);
    u0 = _mm_add_epi16(in0, in1);
    u1 = _mm_add_epi16(in2, in3);
    sum = _mm_add_epi16(sum, u0);

    in0 = _mm_load_si128((const __m128i *)(input + 6 * stride + 0));
    in1 = _mm_load_si128((const __m128i *)(input + 6 * stride + 8));
    in2 = _mm_load_si128((const __m128i *)(input + 7 * stride + 0));
    in3 = _mm_load_si128((const __m128i *)(input + 7 * stride + 8));

    sum = _mm_add_epi16(sum, u1);
    u0 = _mm_add_epi16(in0, in1);
    u1 = _mm_add_epi16(in2, in3);
    sum = _mm_add_epi16(sum, u0);

    sum = _mm_add_epi16(sum, u1);
    input += 8 * stride;
  }

  u0 = _mm_setzero_si128();
  in0 = _mm_unpacklo_epi16(u0, sum);
  in1 = _mm_unpackhi_epi16(u0, sum);
  in0 = _mm_srai_epi32(in0, 16);
  in1 = _mm_srai_epi32(in1, 16);

  sum = _mm_add_epi32(in0, in1);
  in0 = _mm_unpacklo_epi32(sum, u0);
  in1 = _mm_unpackhi_epi32(sum, u0);

  sum = _mm_add_epi32(in0, in1);
  in0 = _mm_srli_si128(sum, 8);

  in1 = _mm_add_epi32(sum, in0);
  in1 = _mm_srai_epi32(in1, 1);
  output[0] = (tran_low_t)_mm_cvtsi128_si32(in1);
}

void aom_fdct32x32_1_sse2(const int16_t *input, tran_low_t *output,
                          int stride) {
  __m128i in0, in1, in2, in3;
  __m128i u0, u1;
  __m128i sum = _mm_setzero_si128();
  int i;

  for (i = 0; i < 8; ++i) {
    in0 = _mm_load_si128((const __m128i *)(input + 0));
    in1 = _mm_load_si128((const __m128i *)(input + 8));
    in2 = _mm_load_si128((const __m128i *)(input + 16));
    in3 = _mm_load_si128((const __m128i *)(input + 24));

    input += stride;
    u0 = _mm_add_epi16(in0, in1);
    u1 = _mm_add_epi16(in2, in3);
    sum = _mm_add_epi16(sum, u0);

    in0 = _mm_load_si128((const __m128i *)(input + 0));
    in1 = _mm_load_si128((const __m128i *)(input + 8));
    in2 = _mm_load_si128((const __m128i *)(input + 16));
    in3 = _mm_load_si128((const __m128i *)(input + 24));

    input += stride;
    sum = _mm_add_epi16(sum, u1);
    u0 = _mm_add_epi16(in0, in1);
    u1 = _mm_add_epi16(in2, in3);
    sum = _mm_add_epi16(sum, u0);

    in0 = _mm_load_si128((const __m128i *)(input + 0));
    in1 = _mm_load_si128((const __m128i *)(input + 8));
    in2 = _mm_load_si128((const __m128i *)(input + 16));
    in3 = _mm_load_si128((const __m128i *)(input + 24));

    input += stride;
    sum = _mm_add_epi16(sum, u1);
    u0 = _mm_add_epi16(in0, in1);
    u1 = _mm_add_epi16(in2, in3);
    sum = _mm_add_epi16(sum, u0);

    in0 = _mm_load_si128((const __m128i *)(input + 0));
    in1 = _mm_load_si128((const __m128i *)(input + 8));
    in2 = _mm_load_si128((const __m128i *)(input + 16));
    in3 = _mm_load_si128((const __m128i *)(input + 24));

    input += stride;
    sum = _mm_add_epi16(sum, u1);
    u0 = _mm_add_epi16(in0, in1);
    u1 = _mm_add_epi16(in2, in3);
    sum = _mm_add_epi16(sum, u0);

    sum = _mm_add_epi16(sum, u1);
  }

  u0 = _mm_setzero_si128();
  in0 = _mm_unpacklo_epi16(u0, sum);
  in1 = _mm_unpackhi_epi16(u0, sum);
  in0 = _mm_srai_epi32(in0, 16);
  in1 = _mm_srai_epi32(in1, 16);

  sum = _mm_add_epi32(in0, in1);
  in0 = _mm_unpacklo_epi32(sum, u0);
  in1 = _mm_unpackhi_epi32(sum, u0);

  sum = _mm_add_epi32(in0, in1);
  in0 = _mm_srli_si128(sum, 8);

  in1 = _mm_add_epi32(sum, in0);
  in1 = _mm_srai_epi32(in1, 3);
  output[0] = (tran_low_t)_mm_cvtsi128_si32(in1);
}

#define DCT_HIGH_BIT_DEPTH 0
#define FDCT4x4_2D aom_fdct4x4_sse2
#define FDCT8x8_2D aom_fdct8x8_sse2
#define FDCT16x16_2D aom_fdct16x16_sse2
#include "aom_dsp/x86/fwd_txfm_impl_sse2.h"
#undef FDCT4x4_2D
#undef FDCT8x8_2D
#undef FDCT16x16_2D

#define FDCT32x32_2D aom_fdct32x32_rd_sse2
#define FDCT32x32_HIGH_PRECISION 0
#include "aom_dsp/x86/fwd_dct32x32_impl_sse2.h"
#undef FDCT32x32_2D
#undef FDCT32x32_HIGH_PRECISION

#define FDCT32x32_2D aom_fdct32x32_sse2
#define FDCT32x32_HIGH_PRECISION 1
#include "aom_dsp/x86/fwd_dct32x32_impl_sse2.h"  // NOLINT
#undef FDCT32x32_2D
#undef FDCT32x32_HIGH_PRECISION
#undef DCT_HIGH_BIT_DEPTH

#if CONFIG_HIGHBITDEPTH
#define DCT_HIGH_BIT_DEPTH 1
#define FDCT4x4_2D aom_highbd_fdct4x4_sse2
#define FDCT8x8_2D aom_highbd_fdct8x8_sse2
#define FDCT16x16_2D aom_highbd_fdct16x16_sse2
#include "aom_dsp/x86/fwd_txfm_impl_sse2.h"  // NOLINT
#undef FDCT4x4_2D
#undef FDCT8x8_2D
#undef FDCT16x16_2D

#define FDCT32x32_2D aom_highbd_fdct32x32_rd_sse2
#define FDCT32x32_HIGH_PRECISION 0
#include "aom_dsp/x86/fwd_dct32x32_impl_sse2.h"  // NOLINT
#undef FDCT32x32_2D
#undef FDCT32x32_HIGH_PRECISION

#define FDCT32x32_2D aom_highbd_fdct32x32_sse2
#define FDCT32x32_HIGH_PRECISION 1
#include "aom_dsp/x86/fwd_dct32x32_impl_sse2.h"  // NOLINT
#undef FDCT32x32_2D
#undef FDCT32x32_HIGH_PRECISION
#undef DCT_HIGH_BIT_DEPTH
#endif  // CONFIG_HIGHBITDEPTH