inv_txfm.c 49.3 KB
Newer Older
1
/*
Yaowu Xu's avatar
Yaowu Xu committed
2
 * Copyright (c) 2016, Alliance for Open Media. All rights reserved
3
 *
Yaowu Xu's avatar
Yaowu Xu committed
4
5
6
7
8
9
 * 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.
10
11
12
13
14
 */

#include <math.h>
#include <string.h>

Yaowu Xu's avatar
Yaowu Xu committed
15
#include "./aom_dsp_rtcd.h"
16
#include "aom_dsp/inv_txfm.h"
17

Yaowu Xu's avatar
Yaowu Xu committed
18
void aom_iwht4x4_16_add_c(const tran_low_t *input, uint8_t *dest, int stride) {
clang-format's avatar
clang-format committed
19
20
  /* 4-point reversible, orthonormal inverse Walsh-Hadamard in 3.5 adds,
     0.5 shifts per pixel. */
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
  int i;
  tran_low_t output[16];
  tran_high_t a1, b1, c1, d1, e1;
  const tran_low_t *ip = input;
  tran_low_t *op = output;

  for (i = 0; i < 4; i++) {
    a1 = ip[0] >> UNIT_QUANT_SHIFT;
    c1 = ip[1] >> UNIT_QUANT_SHIFT;
    d1 = ip[2] >> UNIT_QUANT_SHIFT;
    b1 = ip[3] >> UNIT_QUANT_SHIFT;
    a1 += c1;
    d1 -= b1;
    e1 = (a1 - d1) >> 1;
    b1 = e1 - b1;
    c1 = e1 - c1;
    a1 -= b1;
    d1 += c1;
39
40
41
42
    op[0] = WRAPLOW(a1);
    op[1] = WRAPLOW(b1);
    op[2] = WRAPLOW(c1);
    op[3] = WRAPLOW(d1);
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
    ip += 4;
    op += 4;
  }

  ip = output;
  for (i = 0; i < 4; i++) {
    a1 = ip[4 * 0];
    c1 = ip[4 * 1];
    d1 = ip[4 * 2];
    b1 = ip[4 * 3];
    a1 += c1;
    d1 -= b1;
    e1 = (a1 - d1) >> 1;
    b1 = e1 - b1;
    c1 = e1 - c1;
    a1 -= b1;
    d1 += c1;
60
61
62
63
    dest[stride * 0] = clip_pixel_add(dest[stride * 0], WRAPLOW(a1));
    dest[stride * 1] = clip_pixel_add(dest[stride * 1], WRAPLOW(b1));
    dest[stride * 2] = clip_pixel_add(dest[stride * 2], WRAPLOW(c1));
    dest[stride * 3] = clip_pixel_add(dest[stride * 3], WRAPLOW(d1));
64
65
66
67
68
69

    ip++;
    dest++;
  }
}

Yaowu Xu's avatar
Yaowu Xu committed
70
void aom_iwht4x4_1_add_c(const tran_low_t *in, uint8_t *dest, int dest_stride) {
71
72
73
74
75
76
77
78
79
  int i;
  tran_high_t a1, e1;
  tran_low_t tmp[4];
  const tran_low_t *ip = in;
  tran_low_t *op = tmp;

  a1 = ip[0] >> UNIT_QUANT_SHIFT;
  e1 = a1 >> 1;
  a1 -= e1;
80
81
  op[0] = WRAPLOW(a1);
  op[1] = op[2] = op[3] = WRAPLOW(e1);
82
83
84
85
86
87
88
89
90
91
92
93
94
95

  ip = tmp;
  for (i = 0; i < 4; i++) {
    e1 = ip[0] >> 1;
    a1 = ip[0] - e1;
    dest[dest_stride * 0] = clip_pixel_add(dest[dest_stride * 0], a1);
    dest[dest_stride * 1] = clip_pixel_add(dest[dest_stride * 1], e1);
    dest[dest_stride * 2] = clip_pixel_add(dest[dest_stride * 2], e1);
    dest[dest_stride * 3] = clip_pixel_add(dest[dest_stride * 3], e1);
    ip++;
    dest++;
  }
}

Luca Barbato's avatar
Luca Barbato committed
96
void aom_idct4_c(const tran_low_t *input, tran_low_t *output) {
97
98
99
100
101
  tran_low_t step[4];
  tran_high_t temp1, temp2;
  // stage 1
  temp1 = (input[0] + input[2]) * cospi_16_64;
  temp2 = (input[0] - input[2]) * cospi_16_64;
102
103
  step[0] = WRAPLOW(dct_const_round_shift(temp1));
  step[1] = WRAPLOW(dct_const_round_shift(temp2));
104
105
  temp1 = input[1] * cospi_24_64 - input[3] * cospi_8_64;
  temp2 = input[1] * cospi_8_64 + input[3] * cospi_24_64;
106
107
  step[2] = WRAPLOW(dct_const_round_shift(temp1));
  step[3] = WRAPLOW(dct_const_round_shift(temp2));
108
109

  // stage 2
110
111
112
113
  output[0] = WRAPLOW(step[0] + step[3]);
  output[1] = WRAPLOW(step[1] + step[2]);
  output[2] = WRAPLOW(step[1] - step[2]);
  output[3] = WRAPLOW(step[0] - step[3]);
114
115
}

Yaowu Xu's avatar
Yaowu Xu committed
116
void aom_idct4x4_16_add_c(const tran_low_t *input, uint8_t *dest, int stride) {
117
118
119
120
121
122
123
  tran_low_t out[4 * 4];
  tran_low_t *outptr = out;
  int i, j;
  tran_low_t temp_in[4], temp_out[4];

  // Rows
  for (i = 0; i < 4; ++i) {
Luca Barbato's avatar
Luca Barbato committed
124
    aom_idct4_c(input, outptr);
125
126
127
128
129
130
    input += 4;
    outptr += 4;
  }

  // Columns
  for (i = 0; i < 4; ++i) {
clang-format's avatar
clang-format committed
131
    for (j = 0; j < 4; ++j) temp_in[j] = out[j * 4 + i];
Luca Barbato's avatar
Luca Barbato committed
132
    aom_idct4_c(temp_in, temp_out);
133
134
135
136
137
138
139
    for (j = 0; j < 4; ++j) {
      dest[j * stride + i] = clip_pixel_add(dest[j * stride + i],
                                            ROUND_POWER_OF_TWO(temp_out[j], 4));
    }
  }
}

Yaowu Xu's avatar
Yaowu Xu committed
140
void aom_idct4x4_1_add_c(const tran_low_t *input, uint8_t *dest,
141
142
143
                         int dest_stride) {
  int i;
  tran_high_t a1;
144
145
  tran_low_t out = WRAPLOW(dct_const_round_shift(input[0] * cospi_16_64));
  out = WRAPLOW(dct_const_round_shift(out * cospi_16_64));
146
147
  a1 = ROUND_POWER_OF_TWO(out, 4);

148
149
  if (a1 == 0) return;

150
151
152
153
154
155
156
157
158
  for (i = 0; i < 4; i++) {
    dest[0] = clip_pixel_add(dest[0], a1);
    dest[1] = clip_pixel_add(dest[1], a1);
    dest[2] = clip_pixel_add(dest[2], a1);
    dest[3] = clip_pixel_add(dest[3], a1);
    dest += dest_stride;
  }
}

Luca Barbato's avatar
Luca Barbato committed
159
void aom_idct8_c(const tran_low_t *input, tran_low_t *output) {
160
161
162
163
164
165
166
167
168
  tran_low_t step1[8], step2[8];
  tran_high_t temp1, temp2;
  // stage 1
  step1[0] = input[0];
  step1[2] = input[4];
  step1[1] = input[2];
  step1[3] = input[6];
  temp1 = input[1] * cospi_28_64 - input[7] * cospi_4_64;
  temp2 = input[1] * cospi_4_64 + input[7] * cospi_28_64;
169
170
  step1[4] = WRAPLOW(dct_const_round_shift(temp1));
  step1[7] = WRAPLOW(dct_const_round_shift(temp2));
171
172
  temp1 = input[5] * cospi_12_64 - input[3] * cospi_20_64;
  temp2 = input[5] * cospi_20_64 + input[3] * cospi_12_64;
173
174
  step1[5] = WRAPLOW(dct_const_round_shift(temp1));
  step1[6] = WRAPLOW(dct_const_round_shift(temp2));
175

176
177
178
  // stage 2
  temp1 = (step1[0] + step1[2]) * cospi_16_64;
  temp2 = (step1[0] - step1[2]) * cospi_16_64;
179
180
  step2[0] = WRAPLOW(dct_const_round_shift(temp1));
  step2[1] = WRAPLOW(dct_const_round_shift(temp2));
181
182
  temp1 = step1[1] * cospi_24_64 - step1[3] * cospi_8_64;
  temp2 = step1[1] * cospi_8_64 + step1[3] * cospi_24_64;
183
184
185
186
187
188
  step2[2] = WRAPLOW(dct_const_round_shift(temp1));
  step2[3] = WRAPLOW(dct_const_round_shift(temp2));
  step2[4] = WRAPLOW(step1[4] + step1[5]);
  step2[5] = WRAPLOW(step1[4] - step1[5]);
  step2[6] = WRAPLOW(-step1[6] + step1[7]);
  step2[7] = WRAPLOW(step1[6] + step1[7]);
189

190
  // stage 3
191
192
193
194
  step1[0] = WRAPLOW(step2[0] + step2[3]);
  step1[1] = WRAPLOW(step2[1] + step2[2]);
  step1[2] = WRAPLOW(step2[1] - step2[2]);
  step1[3] = WRAPLOW(step2[0] - step2[3]);
195
196
197
  step1[4] = step2[4];
  temp1 = (step2[6] - step2[5]) * cospi_16_64;
  temp2 = (step2[5] + step2[6]) * cospi_16_64;
198
199
  step1[5] = WRAPLOW(dct_const_round_shift(temp1));
  step1[6] = WRAPLOW(dct_const_round_shift(temp2));
200
201
202
  step1[7] = step2[7];

  // stage 4
203
204
205
206
207
208
209
210
  output[0] = WRAPLOW(step1[0] + step1[7]);
  output[1] = WRAPLOW(step1[1] + step1[6]);
  output[2] = WRAPLOW(step1[2] + step1[5]);
  output[3] = WRAPLOW(step1[3] + step1[4]);
  output[4] = WRAPLOW(step1[3] - step1[4]);
  output[5] = WRAPLOW(step1[2] - step1[5]);
  output[6] = WRAPLOW(step1[1] - step1[6]);
  output[7] = WRAPLOW(step1[0] - step1[7]);
211
212
}

Yaowu Xu's avatar
Yaowu Xu committed
213
void aom_idct8x8_64_add_c(const tran_low_t *input, uint8_t *dest, int stride) {
214
215
216
217
218
219
220
  tran_low_t out[8 * 8];
  tran_low_t *outptr = out;
  int i, j;
  tran_low_t temp_in[8], temp_out[8];

  // First transform rows
  for (i = 0; i < 8; ++i) {
Luca Barbato's avatar
Luca Barbato committed
221
    aom_idct8_c(input, outptr);
222
223
224
225
226
227
    input += 8;
    outptr += 8;
  }

  // Then transform columns
  for (i = 0; i < 8; ++i) {
clang-format's avatar
clang-format committed
228
    for (j = 0; j < 8; ++j) temp_in[j] = out[j * 8 + i];
Luca Barbato's avatar
Luca Barbato committed
229
    aom_idct8_c(temp_in, temp_out);
230
231
232
233
234
235
236
    for (j = 0; j < 8; ++j) {
      dest[j * stride + i] = clip_pixel_add(dest[j * stride + i],
                                            ROUND_POWER_OF_TWO(temp_out[j], 5));
    }
  }
}

Yaowu Xu's avatar
Yaowu Xu committed
237
void aom_idct8x8_1_add_c(const tran_low_t *input, uint8_t *dest, int stride) {
238
239
  int i, j;
  tran_high_t a1;
240
241
  tran_low_t out = WRAPLOW(dct_const_round_shift(input[0] * cospi_16_64));
  out = WRAPLOW(dct_const_round_shift(out * cospi_16_64));
242
  a1 = ROUND_POWER_OF_TWO(out, 5);
243
  if (a1 == 0) return;
244
  for (j = 0; j < 8; ++j) {
clang-format's avatar
clang-format committed
245
    for (i = 0; i < 8; ++i) dest[i] = clip_pixel_add(dest[i], a1);
246
247
248
249
    dest += stride;
  }
}

Luca Barbato's avatar
Luca Barbato committed
250
void aom_iadst4_c(const tran_low_t *input, tran_low_t *output) {
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
  tran_high_t s0, s1, s2, s3, s4, s5, s6, s7;

  tran_low_t x0 = input[0];
  tran_low_t x1 = input[1];
  tran_low_t x2 = input[2];
  tran_low_t x3 = input[3];

  if (!(x0 | x1 | x2 | x3)) {
    output[0] = output[1] = output[2] = output[3] = 0;
    return;
  }

  s0 = sinpi_1_9 * x0;
  s1 = sinpi_2_9 * x0;
  s2 = sinpi_3_9 * x1;
  s3 = sinpi_4_9 * x2;
  s4 = sinpi_1_9 * x2;
  s5 = sinpi_2_9 * x3;
  s6 = sinpi_4_9 * x3;
270
  s7 = WRAPLOW(x0 - x2 + x3);
271
272
273
274
275
276
277
278
279
280

  s0 = s0 + s3 + s5;
  s1 = s1 - s4 - s6;
  s3 = s2;
  s2 = sinpi_3_9 * s7;

  // 1-D transform scaling factor is sqrt(2).
  // The overall dynamic range is 14b (input) + 14b (multiplication scaling)
  // + 1b (addition) = 29b.
  // Hence the output bit depth is 15b.
281
282
283
284
  output[0] = WRAPLOW(dct_const_round_shift(s0 + s3));
  output[1] = WRAPLOW(dct_const_round_shift(s1 + s3));
  output[2] = WRAPLOW(dct_const_round_shift(s2));
  output[3] = WRAPLOW(dct_const_round_shift(s0 + s1 - s3));
285
286
}

Luca Barbato's avatar
Luca Barbato committed
287
void aom_iadst8_c(const tran_low_t *input, tran_low_t *output) {
288
289
290
291
292
293
294
295
296
297
298
299
  int s0, s1, s2, s3, s4, s5, s6, s7;

  tran_high_t x0 = input[7];
  tran_high_t x1 = input[0];
  tran_high_t x2 = input[5];
  tran_high_t x3 = input[2];
  tran_high_t x4 = input[3];
  tran_high_t x5 = input[4];
  tran_high_t x6 = input[1];
  tran_high_t x7 = input[6];

  if (!(x0 | x1 | x2 | x3 | x4 | x5 | x6 | x7)) {
clang-format's avatar
clang-format committed
300
301
    output[0] = output[1] = output[2] = output[3] = output[4] = output[5] =
        output[6] = output[7] = 0;
302
303
304
305
    return;
  }

  // stage 1
clang-format's avatar
clang-format committed
306
307
  s0 = (int)(cospi_2_64 * x0 + cospi_30_64 * x1);
  s1 = (int)(cospi_30_64 * x0 - cospi_2_64 * x1);
308
309
310
311
  s2 = (int)(cospi_10_64 * x2 + cospi_22_64 * x3);
  s3 = (int)(cospi_22_64 * x2 - cospi_10_64 * x3);
  s4 = (int)(cospi_18_64 * x4 + cospi_14_64 * x5);
  s5 = (int)(cospi_14_64 * x4 - cospi_18_64 * x5);
clang-format's avatar
clang-format committed
312
313
  s6 = (int)(cospi_26_64 * x6 + cospi_6_64 * x7);
  s7 = (int)(cospi_6_64 * x6 - cospi_26_64 * x7);
314

315
316
317
318
319
320
321
322
  x0 = WRAPLOW(dct_const_round_shift(s0 + s4));
  x1 = WRAPLOW(dct_const_round_shift(s1 + s5));
  x2 = WRAPLOW(dct_const_round_shift(s2 + s6));
  x3 = WRAPLOW(dct_const_round_shift(s3 + s7));
  x4 = WRAPLOW(dct_const_round_shift(s0 - s4));
  x5 = WRAPLOW(dct_const_round_shift(s1 - s5));
  x6 = WRAPLOW(dct_const_round_shift(s2 - s6));
  x7 = WRAPLOW(dct_const_round_shift(s3 - s7));
323
324
325
326
327
328
329
330
331
332
333

  // stage 2
  s0 = (int)x0;
  s1 = (int)x1;
  s2 = (int)x2;
  s3 = (int)x3;
  s4 = (int)(cospi_8_64 * x4 + cospi_24_64 * x5);
  s5 = (int)(cospi_24_64 * x4 - cospi_8_64 * x5);
  s6 = (int)(-cospi_24_64 * x6 + cospi_8_64 * x7);
  s7 = (int)(cospi_8_64 * x6 + cospi_24_64 * x7);

334
335
336
337
338
339
340
341
  x0 = WRAPLOW(s0 + s2);
  x1 = WRAPLOW(s1 + s3);
  x2 = WRAPLOW(s0 - s2);
  x3 = WRAPLOW(s1 - s3);
  x4 = WRAPLOW(dct_const_round_shift(s4 + s6));
  x5 = WRAPLOW(dct_const_round_shift(s5 + s7));
  x6 = WRAPLOW(dct_const_round_shift(s4 - s6));
  x7 = WRAPLOW(dct_const_round_shift(s5 - s7));
342
343
344
345
346
347
348

  // stage 3
  s2 = (int)(cospi_16_64 * (x2 + x3));
  s3 = (int)(cospi_16_64 * (x2 - x3));
  s6 = (int)(cospi_16_64 * (x6 + x7));
  s7 = (int)(cospi_16_64 * (x6 - x7));

349
350
351
352
353
354
355
356
357
358
359
360
361
  x2 = WRAPLOW(dct_const_round_shift(s2));
  x3 = WRAPLOW(dct_const_round_shift(s3));
  x6 = WRAPLOW(dct_const_round_shift(s6));
  x7 = WRAPLOW(dct_const_round_shift(s7));

  output[0] = WRAPLOW(x0);
  output[1] = WRAPLOW(-x4);
  output[2] = WRAPLOW(x6);
  output[3] = WRAPLOW(-x2);
  output[4] = WRAPLOW(x3);
  output[5] = WRAPLOW(-x7);
  output[6] = WRAPLOW(x5);
  output[7] = WRAPLOW(-x1);
362
363
}

Yaowu Xu's avatar
Yaowu Xu committed
364
void aom_idct8x8_12_add_c(const tran_low_t *input, uint8_t *dest, int stride) {
365
366
367
368
369
370
371
372
  tran_low_t out[8 * 8] = { 0 };
  tran_low_t *outptr = out;
  int i, j;
  tran_low_t temp_in[8], temp_out[8];

  // First transform rows
  // only first 4 row has non-zero coefs
  for (i = 0; i < 4; ++i) {
Luca Barbato's avatar
Luca Barbato committed
373
    aom_idct8_c(input, outptr);
374
375
376
377
378
379
    input += 8;
    outptr += 8;
  }

  // Then transform columns
  for (i = 0; i < 8; ++i) {
clang-format's avatar
clang-format committed
380
    for (j = 0; j < 8; ++j) temp_in[j] = out[j * 8 + i];
Luca Barbato's avatar
Luca Barbato committed
381
    aom_idct8_c(temp_in, temp_out);
382
383
384
385
386
387
388
    for (j = 0; j < 8; ++j) {
      dest[j * stride + i] = clip_pixel_add(dest[j * stride + i],
                                            ROUND_POWER_OF_TWO(temp_out[j], 5));
    }
  }
}

Luca Barbato's avatar
Luca Barbato committed
389
void aom_idct16_c(const tran_low_t *input, tran_low_t *output) {
390
391
392
393
  tran_low_t step1[16], step2[16];
  tran_high_t temp1, temp2;

  // stage 1
clang-format's avatar
clang-format committed
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
  step1[0] = input[0 / 2];
  step1[1] = input[16 / 2];
  step1[2] = input[8 / 2];
  step1[3] = input[24 / 2];
  step1[4] = input[4 / 2];
  step1[5] = input[20 / 2];
  step1[6] = input[12 / 2];
  step1[7] = input[28 / 2];
  step1[8] = input[2 / 2];
  step1[9] = input[18 / 2];
  step1[10] = input[10 / 2];
  step1[11] = input[26 / 2];
  step1[12] = input[6 / 2];
  step1[13] = input[22 / 2];
  step1[14] = input[14 / 2];
  step1[15] = input[30 / 2];
410
411
412
413
414
415
416
417
418
419
420
421
422

  // stage 2
  step2[0] = step1[0];
  step2[1] = step1[1];
  step2[2] = step1[2];
  step2[3] = step1[3];
  step2[4] = step1[4];
  step2[5] = step1[5];
  step2[6] = step1[6];
  step2[7] = step1[7];

  temp1 = step1[8] * cospi_30_64 - step1[15] * cospi_2_64;
  temp2 = step1[8] * cospi_2_64 + step1[15] * cospi_30_64;
423
424
  step2[8] = WRAPLOW(dct_const_round_shift(temp1));
  step2[15] = WRAPLOW(dct_const_round_shift(temp2));
425
426
427

  temp1 = step1[9] * cospi_14_64 - step1[14] * cospi_18_64;
  temp2 = step1[9] * cospi_18_64 + step1[14] * cospi_14_64;
428
429
  step2[9] = WRAPLOW(dct_const_round_shift(temp1));
  step2[14] = WRAPLOW(dct_const_round_shift(temp2));
430
431
432

  temp1 = step1[10] * cospi_22_64 - step1[13] * cospi_10_64;
  temp2 = step1[10] * cospi_10_64 + step1[13] * cospi_22_64;
433
434
  step2[10] = WRAPLOW(dct_const_round_shift(temp1));
  step2[13] = WRAPLOW(dct_const_round_shift(temp2));
435
436
437

  temp1 = step1[11] * cospi_6_64 - step1[12] * cospi_26_64;
  temp2 = step1[11] * cospi_26_64 + step1[12] * cospi_6_64;
438
439
  step2[11] = WRAPLOW(dct_const_round_shift(temp1));
  step2[12] = WRAPLOW(dct_const_round_shift(temp2));
440
441
442
443
444
445
446
447
448

  // stage 3
  step1[0] = step2[0];
  step1[1] = step2[1];
  step1[2] = step2[2];
  step1[3] = step2[3];

  temp1 = step2[4] * cospi_28_64 - step2[7] * cospi_4_64;
  temp2 = step2[4] * cospi_4_64 + step2[7] * cospi_28_64;
449
450
  step1[4] = WRAPLOW(dct_const_round_shift(temp1));
  step1[7] = WRAPLOW(dct_const_round_shift(temp2));
451
452
  temp1 = step2[5] * cospi_12_64 - step2[6] * cospi_20_64;
  temp2 = step2[5] * cospi_20_64 + step2[6] * cospi_12_64;
453
454
455
456
457
458
459
460
461
462
463
  step1[5] = WRAPLOW(dct_const_round_shift(temp1));
  step1[6] = WRAPLOW(dct_const_round_shift(temp2));

  step1[8] = WRAPLOW(step2[8] + step2[9]);
  step1[9] = WRAPLOW(step2[8] - step2[9]);
  step1[10] = WRAPLOW(-step2[10] + step2[11]);
  step1[11] = WRAPLOW(step2[10] + step2[11]);
  step1[12] = WRAPLOW(step2[12] + step2[13]);
  step1[13] = WRAPLOW(step2[12] - step2[13]);
  step1[14] = WRAPLOW(-step2[14] + step2[15]);
  step1[15] = WRAPLOW(step2[14] + step2[15]);
464
465
466
467

  // stage 4
  temp1 = (step1[0] + step1[1]) * cospi_16_64;
  temp2 = (step1[0] - step1[1]) * cospi_16_64;
468
469
  step2[0] = WRAPLOW(dct_const_round_shift(temp1));
  step2[1] = WRAPLOW(dct_const_round_shift(temp2));
470
471
  temp1 = step1[2] * cospi_24_64 - step1[3] * cospi_8_64;
  temp2 = step1[2] * cospi_8_64 + step1[3] * cospi_24_64;
472
473
474
475
476
477
  step2[2] = WRAPLOW(dct_const_round_shift(temp1));
  step2[3] = WRAPLOW(dct_const_round_shift(temp2));
  step2[4] = WRAPLOW(step1[4] + step1[5]);
  step2[5] = WRAPLOW(step1[4] - step1[5]);
  step2[6] = WRAPLOW(-step1[6] + step1[7]);
  step2[7] = WRAPLOW(step1[6] + step1[7]);
478
479
480
481
482

  step2[8] = step1[8];
  step2[15] = step1[15];
  temp1 = -step1[9] * cospi_8_64 + step1[14] * cospi_24_64;
  temp2 = step1[9] * cospi_24_64 + step1[14] * cospi_8_64;
483
484
  step2[9] = WRAPLOW(dct_const_round_shift(temp1));
  step2[14] = WRAPLOW(dct_const_round_shift(temp2));
485
486
  temp1 = -step1[10] * cospi_24_64 - step1[13] * cospi_8_64;
  temp2 = -step1[10] * cospi_8_64 + step1[13] * cospi_24_64;
487
488
  step2[10] = WRAPLOW(dct_const_round_shift(temp1));
  step2[13] = WRAPLOW(dct_const_round_shift(temp2));
489
490
491
492
  step2[11] = step1[11];
  step2[12] = step1[12];

  // stage 5
493
494
495
496
  step1[0] = WRAPLOW(step2[0] + step2[3]);
  step1[1] = WRAPLOW(step2[1] + step2[2]);
  step1[2] = WRAPLOW(step2[1] - step2[2]);
  step1[3] = WRAPLOW(step2[0] - step2[3]);
497
498
499
  step1[4] = step2[4];
  temp1 = (step2[6] - step2[5]) * cospi_16_64;
  temp2 = (step2[5] + step2[6]) * cospi_16_64;
500
501
  step1[5] = WRAPLOW(dct_const_round_shift(temp1));
  step1[6] = WRAPLOW(dct_const_round_shift(temp2));
502
503
  step1[7] = step2[7];

504
505
506
507
508
509
510
511
  step1[8] = WRAPLOW(step2[8] + step2[11]);
  step1[9] = WRAPLOW(step2[9] + step2[10]);
  step1[10] = WRAPLOW(step2[9] - step2[10]);
  step1[11] = WRAPLOW(step2[8] - step2[11]);
  step1[12] = WRAPLOW(-step2[12] + step2[15]);
  step1[13] = WRAPLOW(-step2[13] + step2[14]);
  step1[14] = WRAPLOW(step2[13] + step2[14]);
  step1[15] = WRAPLOW(step2[12] + step2[15]);
512
513

  // stage 6
514
515
516
517
518
519
520
521
  step2[0] = WRAPLOW(step1[0] + step1[7]);
  step2[1] = WRAPLOW(step1[1] + step1[6]);
  step2[2] = WRAPLOW(step1[2] + step1[5]);
  step2[3] = WRAPLOW(step1[3] + step1[4]);
  step2[4] = WRAPLOW(step1[3] - step1[4]);
  step2[5] = WRAPLOW(step1[2] - step1[5]);
  step2[6] = WRAPLOW(step1[1] - step1[6]);
  step2[7] = WRAPLOW(step1[0] - step1[7]);
522
523
524
525
  step2[8] = step1[8];
  step2[9] = step1[9];
  temp1 = (-step1[10] + step1[13]) * cospi_16_64;
  temp2 = (step1[10] + step1[13]) * cospi_16_64;
526
527
  step2[10] = WRAPLOW(dct_const_round_shift(temp1));
  step2[13] = WRAPLOW(dct_const_round_shift(temp2));
528
529
  temp1 = (-step1[11] + step1[12]) * cospi_16_64;
  temp2 = (step1[11] + step1[12]) * cospi_16_64;
530
531
  step2[11] = WRAPLOW(dct_const_round_shift(temp1));
  step2[12] = WRAPLOW(dct_const_round_shift(temp2));
532
533
534
535
  step2[14] = step1[14];
  step2[15] = step1[15];

  // stage 7
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
  output[0] = WRAPLOW(step2[0] + step2[15]);
  output[1] = WRAPLOW(step2[1] + step2[14]);
  output[2] = WRAPLOW(step2[2] + step2[13]);
  output[3] = WRAPLOW(step2[3] + step2[12]);
  output[4] = WRAPLOW(step2[4] + step2[11]);
  output[5] = WRAPLOW(step2[5] + step2[10]);
  output[6] = WRAPLOW(step2[6] + step2[9]);
  output[7] = WRAPLOW(step2[7] + step2[8]);
  output[8] = WRAPLOW(step2[7] - step2[8]);
  output[9] = WRAPLOW(step2[6] - step2[9]);
  output[10] = WRAPLOW(step2[5] - step2[10]);
  output[11] = WRAPLOW(step2[4] - step2[11]);
  output[12] = WRAPLOW(step2[3] - step2[12]);
  output[13] = WRAPLOW(step2[2] - step2[13]);
  output[14] = WRAPLOW(step2[1] - step2[14]);
  output[15] = WRAPLOW(step2[0] - step2[15]);
552
553
}

Yaowu Xu's avatar
Yaowu Xu committed
554
void aom_idct16x16_256_add_c(const tran_low_t *input, uint8_t *dest,
555
556
557
558
559
560
561
562
                             int stride) {
  tran_low_t out[16 * 16];
  tran_low_t *outptr = out;
  int i, j;
  tran_low_t temp_in[16], temp_out[16];

  // First transform rows
  for (i = 0; i < 16; ++i) {
Luca Barbato's avatar
Luca Barbato committed
563
    aom_idct16_c(input, outptr);
564
565
566
567
568
569
    input += 16;
    outptr += 16;
  }

  // Then transform columns
  for (i = 0; i < 16; ++i) {
clang-format's avatar
clang-format committed
570
    for (j = 0; j < 16; ++j) temp_in[j] = out[j * 16 + i];
Luca Barbato's avatar
Luca Barbato committed
571
    aom_idct16_c(temp_in, temp_out);
572
573
574
575
576
577
578
    for (j = 0; j < 16; ++j) {
      dest[j * stride + i] = clip_pixel_add(dest[j * stride + i],
                                            ROUND_POWER_OF_TWO(temp_out[j], 6));
    }
  }
}

Luca Barbato's avatar
Luca Barbato committed
579
void aom_iadst16_c(const tran_low_t *input, tran_low_t *output) {
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
  tran_high_t s0, s1, s2, s3, s4, s5, s6, s7, s8;
  tran_high_t s9, s10, s11, s12, s13, s14, s15;

  tran_high_t x0 = input[15];
  tran_high_t x1 = input[0];
  tran_high_t x2 = input[13];
  tran_high_t x3 = input[2];
  tran_high_t x4 = input[11];
  tran_high_t x5 = input[4];
  tran_high_t x6 = input[9];
  tran_high_t x7 = input[6];
  tran_high_t x8 = input[7];
  tran_high_t x9 = input[8];
  tran_high_t x10 = input[5];
  tran_high_t x11 = input[10];
  tran_high_t x12 = input[3];
  tran_high_t x13 = input[12];
  tran_high_t x14 = input[1];
  tran_high_t x15 = input[14];

clang-format's avatar
clang-format committed
600
601
602
603
604
  if (!(x0 | x1 | x2 | x3 | x4 | x5 | x6 | x7 | x8 | x9 | x10 | x11 | x12 |
        x13 | x14 | x15)) {
    output[0] = output[1] = output[2] = output[3] = output[4] = output[5] =
        output[6] = output[7] = output[8] = output[9] = output[10] =
            output[11] = output[12] = output[13] = output[14] = output[15] = 0;
605
606
607
608
    return;
  }

  // stage 1
clang-format's avatar
clang-format committed
609
  s0 = x0 * cospi_1_64 + x1 * cospi_31_64;
610
  s1 = x0 * cospi_31_64 - x1 * cospi_1_64;
clang-format's avatar
clang-format committed
611
  s2 = x2 * cospi_5_64 + x3 * cospi_27_64;
612
  s3 = x2 * cospi_27_64 - x3 * cospi_5_64;
clang-format's avatar
clang-format committed
613
  s4 = x4 * cospi_9_64 + x5 * cospi_23_64;
614
615
616
617
618
619
620
621
  s5 = x4 * cospi_23_64 - x5 * cospi_9_64;
  s6 = x6 * cospi_13_64 + x7 * cospi_19_64;
  s7 = x6 * cospi_19_64 - x7 * cospi_13_64;
  s8 = x8 * cospi_17_64 + x9 * cospi_15_64;
  s9 = x8 * cospi_15_64 - x9 * cospi_17_64;
  s10 = x10 * cospi_21_64 + x11 * cospi_11_64;
  s11 = x10 * cospi_11_64 - x11 * cospi_21_64;
  s12 = x12 * cospi_25_64 + x13 * cospi_7_64;
clang-format's avatar
clang-format committed
622
  s13 = x12 * cospi_7_64 - x13 * cospi_25_64;
623
  s14 = x14 * cospi_29_64 + x15 * cospi_3_64;
clang-format's avatar
clang-format committed
624
  s15 = x14 * cospi_3_64 - x15 * cospi_29_64;
625

626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
  x0 = WRAPLOW(dct_const_round_shift(s0 + s8));
  x1 = WRAPLOW(dct_const_round_shift(s1 + s9));
  x2 = WRAPLOW(dct_const_round_shift(s2 + s10));
  x3 = WRAPLOW(dct_const_round_shift(s3 + s11));
  x4 = WRAPLOW(dct_const_round_shift(s4 + s12));
  x5 = WRAPLOW(dct_const_round_shift(s5 + s13));
  x6 = WRAPLOW(dct_const_round_shift(s6 + s14));
  x7 = WRAPLOW(dct_const_round_shift(s7 + s15));
  x8 = WRAPLOW(dct_const_round_shift(s0 - s8));
  x9 = WRAPLOW(dct_const_round_shift(s1 - s9));
  x10 = WRAPLOW(dct_const_round_shift(s2 - s10));
  x11 = WRAPLOW(dct_const_round_shift(s3 - s11));
  x12 = WRAPLOW(dct_const_round_shift(s4 - s12));
  x13 = WRAPLOW(dct_const_round_shift(s5 - s13));
  x14 = WRAPLOW(dct_const_round_shift(s6 - s14));
  x15 = WRAPLOW(dct_const_round_shift(s7 - s15));
642
643
644
645
646
647
648
649
650
651

  // stage 2
  s0 = x0;
  s1 = x1;
  s2 = x2;
  s3 = x3;
  s4 = x4;
  s5 = x5;
  s6 = x6;
  s7 = x7;
clang-format's avatar
clang-format committed
652
653
654
655
656
657
658
659
  s8 = x8 * cospi_4_64 + x9 * cospi_28_64;
  s9 = x8 * cospi_28_64 - x9 * cospi_4_64;
  s10 = x10 * cospi_20_64 + x11 * cospi_12_64;
  s11 = x10 * cospi_12_64 - x11 * cospi_20_64;
  s12 = -x12 * cospi_28_64 + x13 * cospi_4_64;
  s13 = x12 * cospi_4_64 + x13 * cospi_28_64;
  s14 = -x14 * cospi_12_64 + x15 * cospi_20_64;
  s15 = x14 * cospi_20_64 + x15 * cospi_12_64;
660

661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
  x0 = WRAPLOW(s0 + s4);
  x1 = WRAPLOW(s1 + s5);
  x2 = WRAPLOW(s2 + s6);
  x3 = WRAPLOW(s3 + s7);
  x4 = WRAPLOW(s0 - s4);
  x5 = WRAPLOW(s1 - s5);
  x6 = WRAPLOW(s2 - s6);
  x7 = WRAPLOW(s3 - s7);
  x8 = WRAPLOW(dct_const_round_shift(s8 + s12));
  x9 = WRAPLOW(dct_const_round_shift(s9 + s13));
  x10 = WRAPLOW(dct_const_round_shift(s10 + s14));
  x11 = WRAPLOW(dct_const_round_shift(s11 + s15));
  x12 = WRAPLOW(dct_const_round_shift(s8 - s12));
  x13 = WRAPLOW(dct_const_round_shift(s9 - s13));
  x14 = WRAPLOW(dct_const_round_shift(s10 - s14));
  x15 = WRAPLOW(dct_const_round_shift(s11 - s15));
677
678
679
680
681
682

  // stage 3
  s0 = x0;
  s1 = x1;
  s2 = x2;
  s3 = x3;
clang-format's avatar
clang-format committed
683
  s4 = x4 * cospi_8_64 + x5 * cospi_24_64;
684
  s5 = x4 * cospi_24_64 - x5 * cospi_8_64;
clang-format's avatar
clang-format committed
685
686
  s6 = -x6 * cospi_24_64 + x7 * cospi_8_64;
  s7 = x6 * cospi_8_64 + x7 * cospi_24_64;
687
688
689
690
  s8 = x8;
  s9 = x9;
  s10 = x10;
  s11 = x11;
clang-format's avatar
clang-format committed
691
  s12 = x12 * cospi_8_64 + x13 * cospi_24_64;
692
  s13 = x12 * cospi_24_64 - x13 * cospi_8_64;
clang-format's avatar
clang-format committed
693
694
  s14 = -x14 * cospi_24_64 + x15 * cospi_8_64;
  s15 = x14 * cospi_8_64 + x15 * cospi_24_64;
695

696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
  x0 = WRAPLOW(s0 + s2);
  x1 = WRAPLOW(s1 + s3);
  x2 = WRAPLOW(s0 - s2);
  x3 = WRAPLOW(s1 - s3);
  x4 = WRAPLOW(dct_const_round_shift(s4 + s6));
  x5 = WRAPLOW(dct_const_round_shift(s5 + s7));
  x6 = WRAPLOW(dct_const_round_shift(s4 - s6));
  x7 = WRAPLOW(dct_const_round_shift(s5 - s7));
  x8 = WRAPLOW(s8 + s10);
  x9 = WRAPLOW(s9 + s11);
  x10 = WRAPLOW(s8 - s10);
  x11 = WRAPLOW(s9 - s11);
  x12 = WRAPLOW(dct_const_round_shift(s12 + s14));
  x13 = WRAPLOW(dct_const_round_shift(s13 + s15));
  x14 = WRAPLOW(dct_const_round_shift(s12 - s14));
  x15 = WRAPLOW(dct_const_round_shift(s13 - s15));
712
713

  // stage 4
clang-format's avatar
clang-format committed
714
  s2 = (-cospi_16_64) * (x2 + x3);
715
716
  s3 = cospi_16_64 * (x2 - x3);
  s6 = cospi_16_64 * (x6 + x7);
clang-format's avatar
clang-format committed
717
  s7 = cospi_16_64 * (-x6 + x7);
718
  s10 = cospi_16_64 * (x10 + x11);
clang-format's avatar
clang-format committed
719
720
  s11 = cospi_16_64 * (-x10 + x11);
  s14 = (-cospi_16_64) * (x14 + x15);
721
722
  s15 = cospi_16_64 * (x14 - x15);

723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
  x2 = WRAPLOW(dct_const_round_shift(s2));
  x3 = WRAPLOW(dct_const_round_shift(s3));
  x6 = WRAPLOW(dct_const_round_shift(s6));
  x7 = WRAPLOW(dct_const_round_shift(s7));
  x10 = WRAPLOW(dct_const_round_shift(s10));
  x11 = WRAPLOW(dct_const_round_shift(s11));
  x14 = WRAPLOW(dct_const_round_shift(s14));
  x15 = WRAPLOW(dct_const_round_shift(s15));

  output[0] = WRAPLOW(x0);
  output[1] = WRAPLOW(-x8);
  output[2] = WRAPLOW(x12);
  output[3] = WRAPLOW(-x4);
  output[4] = WRAPLOW(x6);
  output[5] = WRAPLOW(x14);
  output[6] = WRAPLOW(x10);
  output[7] = WRAPLOW(x2);
  output[8] = WRAPLOW(x3);
  output[9] = WRAPLOW(x11);
  output[10] = WRAPLOW(x15);
  output[11] = WRAPLOW(x7);
  output[12] = WRAPLOW(x5);
  output[13] = WRAPLOW(-x13);
  output[14] = WRAPLOW(x9);
  output[15] = WRAPLOW(-x1);
748
749
}

750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
void aom_idct16x16_38_add_c(const tran_low_t *input, uint8_t *dest,
                            int stride) {
  int i, j;
  tran_low_t out[16 * 16] = { 0 };
  tran_low_t *outptr = out;
  tran_low_t temp_in[16], temp_out[16];

  // First transform rows. Since all non-zero dct coefficients are in
  // upper-left 8x8 area, we only need to calculate first 8 rows here.
  for (i = 0; i < 8; ++i) {
    aom_idct16_c(input, outptr);
    input += 16;
    outptr += 16;
  }

  // Then transform columns
  for (i = 0; i < 16; ++i) {
    for (j = 0; j < 16; ++j) temp_in[j] = out[j * 16 + i];
    aom_idct16_c(temp_in, temp_out);
    for (j = 0; j < 16; ++j) {
      dest[j * stride + i] = clip_pixel_add(dest[j * stride + i],
                                            ROUND_POWER_OF_TWO(temp_out[j], 6));
    }
  }
}

Yaowu Xu's avatar
Yaowu Xu committed
776
void aom_idct16x16_10_add_c(const tran_low_t *input, uint8_t *dest,
777
778
779
780
781
782
783
784
785
                            int stride) {
  tran_low_t out[16 * 16] = { 0 };
  tran_low_t *outptr = out;
  int i, j;
  tran_low_t temp_in[16], temp_out[16];

  // First transform rows. Since all non-zero dct coefficients are in
  // upper-left 4x4 area, we only need to calculate first 4 rows here.
  for (i = 0; i < 4; ++i) {
Luca Barbato's avatar
Luca Barbato committed
786
    aom_idct16_c(input, outptr);
787
788
789
790
791
792
    input += 16;
    outptr += 16;
  }

  // Then transform columns
  for (i = 0; i < 16; ++i) {
clang-format's avatar
clang-format committed
793
    for (j = 0; j < 16; ++j) temp_in[j] = out[j * 16 + i];
Luca Barbato's avatar
Luca Barbato committed
794
    aom_idct16_c(temp_in, temp_out);
795
796
797
798
799
800
801
    for (j = 0; j < 16; ++j) {
      dest[j * stride + i] = clip_pixel_add(dest[j * stride + i],
                                            ROUND_POWER_OF_TWO(temp_out[j], 6));
    }
  }
}

Yaowu Xu's avatar
Yaowu Xu committed
802
void aom_idct16x16_1_add_c(const tran_low_t *input, uint8_t *dest, int stride) {
803
804
  int i, j;
  tran_high_t a1;
805
806
  tran_low_t out = WRAPLOW(dct_const_round_shift(input[0] * cospi_16_64));
  out = WRAPLOW(dct_const_round_shift(out * cospi_16_64));
807
  a1 = ROUND_POWER_OF_TWO(out, 6);
808
  if (a1 == 0) return;
809
  for (j = 0; j < 16; ++j) {
clang-format's avatar
clang-format committed
810
    for (i = 0; i < 16; ++i) dest[i] = clip_pixel_add(dest[i], a1);
811
812
813
814
    dest += stride;
  }
}

Luca Barbato's avatar
Luca Barbato committed
815
void aom_idct32_c(const tran_low_t *input, tran_low_t *output) {
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
  tran_low_t step1[32], step2[32];
  tran_high_t temp1, temp2;

  // stage 1
  step1[0] = input[0];
  step1[1] = input[16];
  step1[2] = input[8];
  step1[3] = input[24];
  step1[4] = input[4];
  step1[5] = input[20];
  step1[6] = input[12];
  step1[7] = input[28];
  step1[8] = input[2];
  step1[9] = input[18];
  step1[10] = input[10];
  step1[11] = input[26];
  step1[12] = input[6];
  step1[13] = input[22];
  step1[14] = input[14];
  step1[15] = input[30];

  temp1 = input[1] * cospi_31_64 - input[31] * cospi_1_64;
  temp2 = input[1] * cospi_1_64 + input[31] * cospi_31_64;
839
840
  step1[16] = WRAPLOW(dct_const_round_shift(temp1));
  step1[31] = WRAPLOW(dct_const_round_shift(temp2));
841
842
843

  temp1 = input[17] * cospi_15_64 - input[15] * cospi_17_64;
  temp2 = input[17] * cospi_17_64 + input[15] * cospi_15_64;
844
845
  step1[17] = WRAPLOW(dct_const_round_shift(temp1));
  step1[30] = WRAPLOW(dct_const_round_shift(temp2));
846
847
848

  temp1 = input[9] * cospi_23_64 - input[23] * cospi_9_64;
  temp2 = input[9] * cospi_9_64 + input[23] * cospi_23_64;
849
850
  step1[18] = WRAPLOW(dct_const_round_shift(temp1));
  step1[29] = WRAPLOW(dct_const_round_shift(temp2));
851
852
853

  temp1 = input[25] * cospi_7_64 - input[7] * cospi_25_64;
  temp2 = input[25] * cospi_25_64 + input[7] * cospi_7_64;
854
855
  step1[19] = WRAPLOW(dct_const_round_shift(temp1));
  step1[28] = WRAPLOW(dct_const_round_shift(temp2));
856
857
858

  temp1 = input[5] * cospi_27_64 - input[27] * cospi_5_64;
  temp2 = input[5] * cospi_5_64 + input[27] * cospi_27_64;
859
860
  step1[20] = WRAPLOW(dct_const_round_shift(temp1));
  step1[27] = WRAPLOW(dct_const_round_shift(temp2));
861
862
863

  temp1 = input[21] * cospi_11_64 - input[11] * cospi_21_64;
  temp2 = input[21] * cospi_21_64 + input[11] * cospi_11_64;
864
865
  step1[21] = WRAPLOW(dct_const_round_shift(temp1));
  step1[26] = WRAPLOW(dct_const_round_shift(temp2));
866
867
868

  temp1 = input[13] * cospi_19_64 - input[19] * cospi_13_64;
  temp2 = input[13] * cospi_13_64 + input[19] * cospi_19_64;
869
870
  step1[22] = WRAPLOW(dct_const_round_shift(temp1));
  step1[25] = WRAPLOW(dct_const_round_shift(temp2));
871
872
873

  temp1 = input[29] * cospi_3_64 - input[3] * cospi_29_64;
  temp2 = input[29] * cospi_29_64 + input[3] * cospi_3_64;
874
875
  step1[23] = WRAPLOW(dct_const_round_shift(temp1));
  step1[24] = WRAPLOW(dct_const_round_shift(temp2));
876
877
878
879
880
881
882
883
884
885
886
887
888

  // stage 2
  step2[0] = step1[0];
  step2[1] = step1[1];
  step2[2] = step1[2];
  step2[3] = step1[3];
  step2[4] = step1[4];
  step2[5] = step1[5];
  step2[6] = step1[6];
  step2[7] = step1[7];

  temp1 = step1[8] * cospi_30_64 - step1[15] * cospi_2_64;
  temp2 = step1[8] * cospi_2_64 + step1[15] * cospi_30_64;
889
890
  step2[8] = WRAPLOW(dct_const_round_shift(temp1));
  step2[15] = WRAPLOW(dct_const_round_shift(temp2));
891
892
893

  temp1 = step1[9] * cospi_14_64 - step1[14] * cospi_18_64;
  temp2 = step1[9] * cospi_18_64 + step1[14] * cospi_14_64;
894
895
  step2[9] = WRAPLOW(dct_const_round_shift(temp1));
  step2[14] = WRAPLOW(dct_const_round_shift(temp2));
896
897
898

  temp1 = step1[10] * cospi_22_64 - step1[13] * cospi_10_64;
  temp2 = step1[10] * cospi_10_64 + step1[13] * cospi_22_64;
899
900
  step2[10] = WRAPLOW(dct_const_round_shift(temp1));
  step2[13] = WRAPLOW(dct_const_round_shift(temp2));
901
902
903

  temp1 = step1[11] * cospi_6_64 - step1[12] * cospi_26_64;
  temp2 = step1[11] * cospi_26_64 + step1[12] * cospi_6_64;
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
  step2[11] = WRAPLOW(dct_const_round_shift(temp1));
  step2[12] = WRAPLOW(dct_const_round_shift(temp2));

  step2[16] = WRAPLOW(step1[16] + step1[17]);
  step2[17] = WRAPLOW(step1[16] - step1[17]);
  step2[18] = WRAPLOW(-step1[18] + step1[19]);
  step2[19] = WRAPLOW(step1[18] + step1[19]);
  step2[20] = WRAPLOW(step1[20] + step1[21]);
  step2[21] = WRAPLOW(step1[20] - step1[21]);
  step2[22] = WRAPLOW(-step1[22] + step1[23]);
  step2[23] = WRAPLOW(step1[22] + step1[23]);
  step2[24] = WRAPLOW(step1[24] + step1[25]);
  step2[25] = WRAPLOW(step1[24] - step1[25]);
  step2[26] = WRAPLOW(-step1[26] + step1[27]);
  step2[27] = WRAPLOW(step1[26] + step1[27]);
  step2[28] = WRAPLOW(step1[28] + step1[29]);
  step2[29] = WRAPLOW(step1[28] - step1[29]);
  step2[30] = WRAPLOW(-step1[30] + step1[31]);
  step2[31] = WRAPLOW(step1[30] + step1[31]);
923
924
925
926
927
928
929
930
931

  // stage 3
  step1[0] = step2[0];
  step1[1] = step2[1];
  step1[2] = step2[2];
  step1[3] = step2[3];

  temp1 = step2[4] * cospi_28_64 - step2[7] * cospi_4_64;
  temp2 = step2[4] * cospi_4_64 + step2[7] * cospi_28_64;
932
933
  step1[4] = WRAPLOW(dct_const_round_shift(temp1));
  step1[7] = WRAPLOW(dct_const_round_shift(temp2));
934
935
  temp1 = step2[5] * cospi_12_64 - step2[6] * cospi_20_64;
  temp2 = step2[5] * cospi_20_64 + step2[6] * cospi_12_64;
936
937
938
939
940
941
942
943
944
945
946
  step1[5] = WRAPLOW(dct_const_round_shift(temp1));
  step1[6] = WRAPLOW(dct_const_round_shift(temp2));

  step1[8] = WRAPLOW(step2[8] + step2[9]);
  step1[9] = WRAPLOW(step2[8] - step2[9]);
  step1[10] = WRAPLOW(-step2[10] + step2[11]);
  step1[11] = WRAPLOW(step2[10] + step2[11]);
  step1[12] = WRAPLOW(step2[12] + step2[13]);
  step1[13] = WRAPLOW(step2[12] - step2[13]);
  step1[14] = WRAPLOW(-step2[14] + step2[15]);
  step1[15] = WRAPLOW(step2[14] + step2[15]);
947
948
949
950
951

  step1[16] = step2[16];
  step1[31] = step2[31];
  temp1 = -step2[17] * cospi_4_64 + step2[30] * cospi_28_64;
  temp2 = step2[17] * cospi_28_64 + step2[30] * cospi_4_64;
952
953
  step1[17] = WRAPLOW(dct_const_round_shift(temp1));
  step1[30] = WRAPLOW(dct_const_round_shift(temp2));
954
955
  temp1 = -step2[18] * cospi_28_64 - step2[29] * cospi_4_64;
  temp2 = -step2[18] * cospi_4_64 + step2[29] * cospi_28_64;
956
957
  step1[18] = WRAPLOW(dct_const_round_shift(temp1));
  step1[29] = WRAPLOW(dct_const_round_shift(temp2));
958
959
960
961
  step1[19] = step2[19];
  step1[20] = step2[20];
  temp1 = -step2[21] * cospi_20_64 + step2[26] * cospi_12_64;
  temp2 = step2[21] * cospi_12_64 + step2[26] * cospi_20_64;
962
963
  step1[21] = WRAPLOW(dct_const_round_shift(temp1));
  step1[26] = WRAPLOW(dct_const_round_shift(temp2));
964
965
  temp1 = -step2[22] * cospi_12_64 - step2[25] * cospi_20_64;
  temp2 = -step2[22] * cospi_20_64 + step2[25] * cospi_12_64;
966
967
  step1[22] = WRAPLOW(dct_const_round_shift(temp1));
  step1[25] = WRAPLOW(dct_const_round_shift(temp2));
968
969
970
971
972
973
974
975
  step1[23] = step2[23];
  step1[24] = step2[24];
  step1[27] = step2[27];
  step1[28] = step2[28];

  // stage 4
  temp1 = (step1[0] + step1[1]) * cospi_16_64;
  temp2 = (step1[0] - step1[1]) * cospi_16_64;
976
977
  step2[0] = WRAPLOW(dct_const_round_shift(temp1));
  step2[1] = WRAPLOW(dct_const_round_shift(temp2));
978
979
  temp1 = step1[2] * cospi_24_64 - step1[3] * cospi_8_64;
  temp2 = step1[2] * cospi_8_64 + step1[3] * cospi_24_64;
980
981
982
983
984
985
  step2[2] = WRAPLOW(dct_const_round_shift(temp1));
  step2[3] = WRAPLOW(dct_const_round_shift(temp2));
  step2[4] = WRAPLOW(step1[4] + step1[5]);
  step2[5] = WRAPLOW(step1[4] - step1[5]);
  step2[6] = WRAPLOW(-step1[6] + step1[7]);
  step2[7] = WRAPLOW(step1[6] + step1[7]);
986
987
988
989
990

  step2[8] = step1[8];
  step2[15] = step1[15];
  temp1 = -step1[9] * cospi_8_64 + step1[14] * cospi_24_64;
  temp2 = step1[9] * cospi_24_64 + step1[14] * cospi_8_64;
991
992
  step2[9] = WRAPLOW(dct_const_round_shift(temp1));
  step2[14] = WRAPLOW(dct_const_round_shift(temp2));
993
994
  temp1 = -step1[10] * cospi_24_64 - step1[13] * cospi_8_64;
  temp2 = -step1[10] * cospi_8_64 + step1[13] * cospi_24_64;
995
996
  step2[10] = WRAPLOW(dct_const_round_shift(temp1));
  step2[13] = WRAPLOW(dct_const_round_shift(temp2));
997
998
999
  step2[11] = step1[11];
  step2[12] = step1[12];

1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
  step2[16] = WRAPLOW(step1[16] + step1[19]);
  step2[17] = WRAPLOW(step1[17] + step1[18]);
  step2[18] = WRAPLOW(step1[17] - step1[18]);
  step2[19] = WRAPLOW(step1[16] - step1[19]);
  step2[20] = WRAPLOW(-step1[20] + step1[23]);
  step2[21] = WRAPLOW(-step1[21] + step1[22]);
  step2[22] = WRAPLOW(step1[21] + step1[22]);
  step2[23] = WRAPLOW(step1[20] + step1[23]);

  step2[24] = WRAPLOW(step1[24] + step1[27]);
  step2[25] = WRAPLOW(step1[25] + step1[26]);
  step2[26] = WRAPLOW(step1[25] - step1[26]);
  step2[27] = WRAPLOW(step1[24] - step1[27]);
  step2[28] = WRAPLOW(-step1[28] + step1[31]);
  step2[29] = WRAPLOW(-step1[29] + step1[30]);
  step2[30] = WRAPLOW(step1[29] + step1[30]);
  step2[31] = WRAPLOW(step1[28] + step1[31]);
1017
1018

  // stage 5
1019
1020
1021
1022
  step1[0] = WRAPLOW(step2[0] + step2[3]);
  step1[1] = WRAPLOW(step2[1] + step2[2]);
  step1[2] = WRAPLOW(step2[1] - step2[2]);
  step1[3] = WRAPLOW(step2[0] - step2[3]);
1023
1024
1025
  step1[4] = step2[4];
  temp1 = (step2[6] - step2[5]) * cospi_16_64;
  temp2 = (step2[5] + step2[6]) * cospi_16_64;
1026
1027
  step1[5] = WRAPLOW(dct_const_round_shift(temp1));
  step1[6] = WRAPLOW(dct_const_round_shift(temp2));
1028
1029
  step1[7] = step2[7];

1030
1031
1032
1033
1034
1035
1036
1037
  step1[8] = WRAPLOW(step2[8] + step2[11]);
  step1[9] = WRAPLOW(step2[9] + step2[10]);
  step1[10] = WRAPLOW(step2[9] - step2[10]);
  step1[11] = WRAPLOW(step2[8] - step2[11]);
  step1[12] = WRAPLOW(-step2[12] + step2[15]);
  step1[13] = WRAPLOW(-step2[13] + step2[14]);
  step1[14] = WRAPLOW(step2[13] + step2[14]);
  step1[15] = WRAPLOW(step2[12] + step2[15]);
1038
1039
1040
1041
1042

  step1[16] = step2[16];
  step1[17] = step2[17];
  temp1 = -step2[18] * cospi_8_64 + step2[29] * cospi_24_64;
  temp2 = step2[18] * cospi_24_64 + step2[29] * cospi_8_64;
1043
1044
  step1[18] = WRAPLOW(dct_const_round_shift(temp1));
  step1[29] = WRAPLOW(dct_const_round_shift(temp2));
1045
1046
  temp1 = -step2[19] * cospi_8_64 + step2[28] * cospi_24_64;
  temp2 = step2[19] * cospi_24_64 + step2[28] * cospi_8_64;
1047
1048
  step1[19] = WRAPLOW(dct_const_round_shift(temp1));
  step1[28] = WRAPLOW(dct_const_round_shift(temp2));
1049
1050
  temp1 = -step2[20] * cospi_24_64 - step2[27] * cospi_8_64;
  temp2 = -step2[20] * cospi_8_64 + step2[27] * cospi_24_64;
1051
1052
  step1[20] = WRAPLOW(dct_const_round_shift(temp1));
  step1[27] = WRAPLOW(dct_const_round_shift(temp2));
1053
1054
  temp1 = -step2[21] * cospi_24_64 - step2[26] * cospi_8_64;
  temp2 = -step2[21] * cospi_8_64 + step2[26] * cospi_24_64;
1055
1056
  step1[21] = WRAPLOW(dct_const_round_shift(temp1));
  step1[26] = WRAPLOW(dct_const_round_shift(temp2));
1057
1058
1059
1060
1061
1062
1063
1064
  step1[22] = step2[22];
  step1[23] = step2[23];
  step1[24] = step2[24];
  step1[25] = step2[25];
  step1[30] = step2[30];
  step1[31] = step2[31];

  // stage 6
1065
1066
1067
1068
1069
1070
1071
1072
  step2[0] = WRAPLOW(step1[0] + step1[7]);
  step2[1] = WRAPLOW(step1[1] + step1[6]);
  step2[2] = WRAPLOW(step1[2] + step1[5]);
  step2[3] = WRAPLOW(step1[3] + step1[4]);
  step2[4] = WRAPLOW(step1[3] - step1[4]);
  step2[5] = WRAPLOW(step1[2] - step1[5]);
  step2[6] = WRAPLOW(step1[1] - step1[6]);
  step2[7] = WRAPLOW(step1[0] - step1[7]);
1073
1074
1075
1076
  step2[8] = step1[8];
  step2[9] = step1[9];
  temp1 = (-step1[10] + step1[13]) * cospi_16_64;
  temp2 = (step1[10] + step1[13]) * cospi_16_64;
1077
1078
  step2[10] = WRAPLOW(dct_const_round_shift(temp1));
  step2[13] = WRAPLOW(dct_const_round_shift(temp2));
1079
1080
  temp1 = (-step1[11] + step1[12]) * cospi_16_64;
  temp2 = (step1[11] + step1[12]) * cospi_16_64;
1081
1082
  step2[11] = WRAPLOW(dct_const_round_shift(temp1));
  step2[12] = WRAPLOW(dct_const_round_shift(temp2));
1083
1084
1085
  step2[14] = step1[14];
  step2[15] = step1[15];

1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
  step2[16] = WRAPLOW(step1[16] + step1[23]);
  step2[17] = WRAPLOW(step1[17] + step1[22]);
  step2[18] = WRAPLOW(step1[18] + step1[21]);
  step2[19] = WRAPLOW(step1[19] + step1[20]);
  step2[20] = WRAPLOW(step1[19] - step1[20]);
  step2[21] = WRAPLOW(step1[18] - step1[21]);
  step2[22] = WRAPLOW(step1[17] - step1[22]);
  step2[23] = WRAPLOW(step1[16] - step1[23]);

  step2[24] = WRAPLOW(-step1[24] + step1[31]);
  step2[25] = WRAPLOW(-step1[25] + step1[30]);
  step2[26] = WRAPLOW(-step1[26] + step1[29]);
  step2[27] = WRAPLOW(-step1[27] + step1[28]);
  step2[28] = WRAPLOW(step1[27] + step1[28]);
  step2[29] = WRAPLOW(step1[26] + step1[29]);
  step2[30] = WRAPLOW(step1[25] + step1[30]);
  step2[31] = WRAPLOW(step1[24] + step1[31]);
1103
1104

  // stage 7
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
  step1[0] = WRAPLOW(step2[0] + step2[15]);
  step1[1] = WRAPLOW(step2[1] + step2[14]);
  step1[2] = WRAPLOW(step2[2] + step2[13]);
  step1[3] = WRAPLOW(step2[3] + step2[12]);
  step1[4] = WRAPLOW(step2[4] + step2[11]);
  step1[5] = WRAPLOW(step2[5] + step2[10]);
  step1[6] = WRAPLOW(step2[6] + step2[9]);
  step1[7] = WRAPLOW(step2[7] + step2[8]);
  step1[8] = WRAPLOW(step2[7] - step2[8]);
  step1[9] = WRAPLOW(step2[6] - step2[9]);
  step1[10] = WRAPLOW(step2[5] - step2[10]);
  step1[11] = WRAPLOW(step2[4] - step2[11]);
  step1[12] = WRAPLOW(step2[3] - step2[12]);
  step1[13] = WRAPLOW(step2[2] - step2[13]);
  step1[14] = WRAPLOW(step2[1] - step2[14]);
  step1[15] = WRAPLOW(step2[0] - step2[15]);
1121
1122
1123
1124
1125
1126
1127

  step1[16] = step2[16];
  step1[17] = step2[17];
  step1[18] = step2[18];
  step1[19] = step2[19];
  temp1 = (-step2[20] + step2[27]) * cospi_16_64;
  temp2 = (step2[20] + step2[27]) * cospi_16_64;
1128
1129
  step1[20] = WRAPLOW(dct_const_round_shift(temp1));
  step1[27] = WRAPLOW(dct_const_round_shift(temp2));
1130
1131
  temp1 = (-step2[21] + step2[26]) * cospi_16_64;
  temp2 = (step2[21] + step2[26]) * cospi_16_64;
1132
1133
  step1[21] = WRAPLOW(dct_const_round_shift(temp1));
  step1[26] = WRAPLOW(dct_const_round_shift(temp2));
1134
1135
  temp1 = (-step2[22] + step2[25]) * cospi_16_64;
  temp2 = (step2[22] + step2[25]) * cospi_16_64;
1136
1137
  step1[22] = WRAPLOW(dct_const_round_shift(temp1));
  step1[25] = WRAPLOW(dct_const_round_shift(temp2));
1138
1139
  temp1 = (-step2[23] + step2[24]) * cospi_16_64;
  temp2 = (step2[23] + step2[24]) * cospi_16_64;
1140
1141
  step1[23] = WRAPLOW(dct_const_round_shift(temp1));
  step1[24] = WRAPLOW(dct_const_round_shift(temp2));
1142
1143
1144
1145
1146
1147
  step1[28] = step2[28];
  step1[29] = step2[29];
  step1[30] = step2[30];
  step1[31] = step2[31];

  // final stage
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
  output[0] = WRAPLOW(step1[0] + step1[31]);
  output[1] = WRAPLOW(step1[1] + step1[30]);
  output[2] = WRAPLOW(step1[2] + step1[29]);
  output[3] = WRAPLOW(step1[3] + step1[28]);
  output[4] = WRAPLOW(step1[4] + step1[27]);
  output[5] = WRAPLOW(step1[5] + step1[26]);
  output[6] = WRAPLOW(step1[6] + step1[25]);
  output[7] = WRAPLOW(step1[7] + step1[24]);
  output[8] = WRAPLOW(step1[8] + step1[23]);
  output[9] = WRAPLOW(step1[9] + step1[22]);
  output[10] = WRAPLOW(step1[10] + step1[21]);
  output[11] = WRAPLOW(step1[11] + step1[20]);
  output[12] = WRAPLOW(step1[12] + step1[19]);
  output[13] = WRAPLOW(step1[13] + step1[18]);
  output[14] = WRAPLOW(step1[14] + step1[17]);
  output[15] = WRAPLOW(step1[15] + step1[16]);
  output[16] = WRAPLOW(step1[15] - step1[16]);
  output[17] = WRAPLOW(step1[14] - step1[17]);
  output[18] = WRAPLOW(step1[13] - step1[18]);
  output[19] = WRAPLOW(step1[12] - step1[19]);
  output[20] = WRAPLOW(step1[11] - step1[20]);
  output[21] = WRAPLOW(step1[10] - step1[21]);
  output[22] = WRAPLOW(step1[9] - step1[22]);
  output[23] = WRAPLOW(step1[8] - step1[23]);
  output[24] = WRAPLOW(step1[7] - step1[24]);
  output[25] = WRAPLOW(step1[6] - step1[25]);
  output[26] = WRAPLOW(step1[5] - step1[26]);
  output[27] = WRAPLOW(step1[4] - step1[27]);
  output[28] = WRAPLOW(step1[3] - step1[28]);
  output[29] = WRAPLOW(step1[2] - step1[29]);
  output[30] = WRAPLOW(step1[1] - step1[30]);
  output[31] = WRAPLOW(step1[0] - step1[31]);
1180
1181
}

1182
#if CONFIG_MRC_TX
Sarah Parker's avatar
Sarah Parker committed
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
void aom_imrc32x32_1_add_c(const tran_low_t *input, uint8_t *dest, int stride,
                           uint8_t *mask) {
  int i, j;
  tran_high_t a1;

  tran_low_t out = WRAPLOW(dct_const_round_shift(input[0] * cospi_16_64));
  out = WRAPLOW(dct_const_round_shift(out * cospi_16_64));
  a1 = ROUND_POWER_OF_TWO(out, 6);
  if (a1 == 0) return;

  for (j = 0; j < 32; ++j) {
    for (i = 0; i < 32; ++i) {
      int mask_val = mask[i];
      dest[i] = mask_val ? clip_pixel_add(dest[i], a1) : dest[i];
    }
    mask += 32;
    dest += stride;
  }
}

1203
void aom_imrc32x32_1024_add_c(const tran_low_t *input, uint8_t *dest,
Sarah Parker's avatar
Sarah Parker committed
1204
                              int stride, uint8_t *mask) {
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
  tran_low_t out[32 * 32];
  tran_low_t *outptr = out;
  int i, j;
  tran_low_t temp_in[32], temp_out[32];

  // Rows
  for (i = 0; i < 32; ++i) {
    int16_t zero_coeff[16];
    for (j = 0; j < 16; ++j) zero_coeff[j] = input[2 * j] | input[2 * j + 1];
    for (j = 0; j < 8; ++j)
      zero_coeff[j] = zero_coeff[2 * j] | zero_coeff[2 * j + 1];
    for (j = 0; j < 4; ++j)
      zero_coeff[j] = zero_coeff[2 * j] | zero_coeff[2 * j + 1];
    for (j = 0; j < 2; ++j)
      zero_coeff[j] = zero_coeff[2 * j] | zero_coeff[2 * j + 1];

    if (zero_coeff[0] | zero_coeff[1])
      aom_idct32_c(input, outptr);
    else
      memset(outptr, 0, sizeof(tran_low_t) * 32);
    input += 32;
    outptr += 32;
  }

  // Columns
  for (i = 0; i < 32; ++i) {
    for (j = 0; j < 32; ++j) temp_in[j] = out[j * 32 + i];
    aom_idct32_c(temp_in, temp_out);
    for (j = 0; j < 32; ++j) {
      // Only add the coefficient if the mask value is 1
      int mask_val = mask[j * 32 + i];
      dest[j * stride + i] =
          mask_val ? clip_pixel_add(dest[j * stride + i],
                                    ROUND_POWER_OF_TWO(temp_out[j], 6))
                   : dest[j * stride + i];
    }
  }
}