od_dering_simd.h 13.1 KB
Newer Older
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
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
/*
 * 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 "./av1_rtcd.h"
#include "./od_dering.h"

/* partial A is a 16-bit vector of the form:
   [x8 x7 x6 x5 x4 x3 x2 x1] and partial B has the form:
   [0  y1 y2 y3 y4 y5 y6 y7].
   This function computes (x1^2+y1^2)*C1 + (x2^2+y2^2)*C2 + ...
   (x7^2+y2^7)*C7 + (x8^2+0^2)*C8 where the C1..C8 constants are in const1
   and const2. */
static INLINE v128 fold_mul_and_sum(v128 partiala, v128 partialb, v128 const1,
                                    v128 const2) {
  v128 tmp;
  /* Reverse partial B. */
  partialb = v128_shuffle_8(
      partialb, v128_from_32(0x0f0e0100, 0x03020504, 0x07060908, 0x0b0a0d0c));
  /* Interleave the x and y values of identical indices and pair x8 with 0. */
  tmp = partiala;
  partiala = v128_ziplo_16(partialb, partiala);
  partialb = v128_ziphi_16(partialb, tmp);
  /* Square and add the corresponding x and y values. */
  partiala = v128_madd_s16(partiala, partiala);
  partialb = v128_madd_s16(partialb, partialb);
  /* Multiply by constant. */
  partiala = v128_mullo_s32(partiala, const1);
  partialb = v128_mullo_s32(partialb, const2);
  /* Sum all results. */
  partiala = v128_add_32(partiala, partialb);
  return partiala;
}

static INLINE v128 hsum4(v128 x0, v128 x1, v128 x2, v128 x3) {
  v128 t0, t1, t2, t3;
  t0 = v128_ziplo_32(x1, x0);
  t1 = v128_ziplo_32(x3, x2);
  t2 = v128_ziphi_32(x1, x0);
  t3 = v128_ziphi_32(x3, x2);
  x0 = v128_ziplo_64(t1, t0);
  x1 = v128_ziphi_64(t1, t0);
  x2 = v128_ziplo_64(t3, t2);
  x3 = v128_ziphi_64(t3, t2);
  return v128_add_32(v128_add_32(x0, x1), v128_add_32(x2, x3));
}

/* Computes cost for directions 0, 5, 6 and 7. We can call this function again
   to compute the remaining directions. */
static INLINE void compute_directions(v128 lines[8], int32_t tmp_cost1[4]) {
  v128 partial4a, partial4b, partial5a, partial5b, partial7a, partial7b;
  v128 partial6;
  v128 tmp;
  /* Partial sums for lines 0 and 1. */
  partial4a = v128_shl_n_byte(lines[0], 14);
  partial4b = v128_shr_n_byte(lines[0], 2);
  partial4a = v128_add_16(partial4a, v128_shl_n_byte(lines[1], 12));
  partial4b = v128_add_16(partial4b, v128_shr_n_byte(lines[1], 4));
  tmp = v128_add_16(lines[0], lines[1]);
  partial5a = v128_shl_n_byte(tmp, 10);
  partial5b = v128_shr_n_byte(tmp, 6);
  partial7a = v128_shl_n_byte(tmp, 4);
  partial7b = v128_shr_n_byte(tmp, 12);
  partial6 = tmp;

  /* Partial sums for lines 2 and 3. */
  partial4a = v128_add_16(partial4a, v128_shl_n_byte(lines[2], 10));
  partial4b = v128_add_16(partial4b, v128_shr_n_byte(lines[2], 6));
  partial4a = v128_add_16(partial4a, v128_shl_n_byte(lines[3], 8));
  partial4b = v128_add_16(partial4b, v128_shr_n_byte(lines[3], 8));
  tmp = v128_add_16(lines[2], lines[3]);
  partial5a = v128_add_16(partial5a, v128_shl_n_byte(tmp, 8));
  partial5b = v128_add_16(partial5b, v128_shr_n_byte(tmp, 8));
  partial7a = v128_add_16(partial7a, v128_shl_n_byte(tmp, 6));
  partial7b = v128_add_16(partial7b, v128_shr_n_byte(tmp, 10));
  partial6 = v128_add_16(partial6, tmp);

  /* Partial sums for lines 4 and 5. */
  partial4a = v128_add_16(partial4a, v128_shl_n_byte(lines[4], 6));
  partial4b = v128_add_16(partial4b, v128_shr_n_byte(lines[4], 10));
  partial4a = v128_add_16(partial4a, v128_shl_n_byte(lines[5], 4));
  partial4b = v128_add_16(partial4b, v128_shr_n_byte(lines[5], 12));
  tmp = v128_add_16(lines[4], lines[5]);
  partial5a = v128_add_16(partial5a, v128_shl_n_byte(tmp, 6));
  partial5b = v128_add_16(partial5b, v128_shr_n_byte(tmp, 10));
  partial7a = v128_add_16(partial7a, v128_shl_n_byte(tmp, 8));
  partial7b = v128_add_16(partial7b, v128_shr_n_byte(tmp, 8));
  partial6 = v128_add_16(partial6, tmp);

  /* Partial sums for lines 6 and 7. */
  partial4a = v128_add_16(partial4a, v128_shl_n_byte(lines[6], 2));
  partial4b = v128_add_16(partial4b, v128_shr_n_byte(lines[6], 14));
  partial4a = v128_add_16(partial4a, lines[7]);
  tmp = v128_add_16(lines[6], lines[7]);
  partial5a = v128_add_16(partial5a, v128_shl_n_byte(tmp, 4));
  partial5b = v128_add_16(partial5b, v128_shr_n_byte(tmp, 12));
  partial7a = v128_add_16(partial7a, v128_shl_n_byte(tmp, 10));
  partial7b = v128_add_16(partial7b, v128_shr_n_byte(tmp, 6));
  partial6 = v128_add_16(partial6, tmp);

  /* Compute costs in terms of partial sums. */
  partial4a =
      fold_mul_and_sum(partial4a, partial4b, v128_from_32(210, 280, 420, 840),
                       v128_from_32(105, 120, 140, 168));
  partial7a =
      fold_mul_and_sum(partial7a, partial7b, v128_from_32(210, 420, 0, 0),
                       v128_from_32(105, 105, 105, 140));
  partial5a =
      fold_mul_and_sum(partial5a, partial5b, v128_from_32(210, 420, 0, 0),
                       v128_from_32(105, 105, 105, 140));
  partial6 = v128_madd_s16(partial6, partial6);
  partial6 = v128_mullo_s32(partial6, v128_dup_32(105));

  partial4a = hsum4(partial4a, partial5a, partial6, partial7a);
  v128_store_unaligned(tmp_cost1, partial4a);
}

/* transpose and reverse the order of the lines -- equivalent to a 90-degree
   counter-clockwise rotation of the pixels. */
static INLINE void array_reverse_transpose_8x8(v128 *in, v128 *res) {
  const v128 tr0_0 = v128_ziplo_16(in[1], in[0]);
  const v128 tr0_1 = v128_ziplo_16(in[3], in[2]);
  const v128 tr0_2 = v128_ziphi_16(in[1], in[0]);
  const v128 tr0_3 = v128_ziphi_16(in[3], in[2]);
  const v128 tr0_4 = v128_ziplo_16(in[5], in[4]);
  const v128 tr0_5 = v128_ziplo_16(in[7], in[6]);
  const v128 tr0_6 = v128_ziphi_16(in[5], in[4]);
  const v128 tr0_7 = v128_ziphi_16(in[7], in[6]);

  const v128 tr1_0 = v128_ziplo_32(tr0_1, tr0_0);
  const v128 tr1_1 = v128_ziplo_32(tr0_5, tr0_4);
  const v128 tr1_2 = v128_ziphi_32(tr0_1, tr0_0);
  const v128 tr1_3 = v128_ziphi_32(tr0_5, tr0_4);
  const v128 tr1_4 = v128_ziplo_32(tr0_3, tr0_2);
  const v128 tr1_5 = v128_ziplo_32(tr0_7, tr0_6);
  const v128 tr1_6 = v128_ziphi_32(tr0_3, tr0_2);
  const v128 tr1_7 = v128_ziphi_32(tr0_7, tr0_6);

  res[7] = v128_ziplo_64(tr1_1, tr1_0);
  res[6] = v128_ziphi_64(tr1_1, tr1_0);
  res[5] = v128_ziplo_64(tr1_3, tr1_2);
  res[4] = v128_ziphi_64(tr1_3, tr1_2);
  res[3] = v128_ziplo_64(tr1_5, tr1_4);
  res[2] = v128_ziphi_64(tr1_5, tr1_4);
  res[1] = v128_ziplo_64(tr1_7, tr1_6);
  res[0] = v128_ziphi_64(tr1_7, tr1_6);
}

int SIMD_FUNC(od_dir_find8)(const od_dering_in *img, int stride, int32_t *var,
                            int coeff_shift) {
  int i;
  int32_t cost[8];
  int32_t best_cost = 0;
  int best_dir = 0;
  v128 lines[8];
  for (i = 0; i < 8; i++) {
    lines[i] = v128_load_unaligned(&img[i * stride]);
    lines[i] =
        v128_sub_16(v128_shr_s16(lines[i], coeff_shift), v128_dup_16(128));
  }

  /* Compute "mostly vertical" directions. */
  compute_directions(lines, cost + 4);

  array_reverse_transpose_8x8(lines, lines);

  /* Compute "mostly horizontal" directions. */
  compute_directions(lines, cost);

  for (i = 0; i < 8; i++) {
    if (cost[i] > best_cost) {
      best_cost = cost[i];
      best_dir = i;
    }
  }

  /* Difference between the optimal variance and the variance along the
     orthogonal direction. Again, the sum(x^2) terms cancel out. */
  *var = best_cost - cost[(best_dir + 4) & 7];
  /* We'd normally divide by 840, but dividing by 1024 is close enough
     for what we're going to do with this. */
  *var >>= 10;
  return best_dir;
}

static INLINE v128 od_cmplt_abs_epi16(v128 in, v128 threshold) {
  return v128_cmplt_s16(v128_abs_s16(in), threshold);
}

int SIMD_FUNC(od_filter_dering_direction_4x4)(uint16_t *y, int ystride,
                                              const uint16_t *in, int threshold,
                                              int dir) {
  int i;
  v128 sum;
  v128 p;
  v128 cmp;
  v128 row;
  v128 res;
  v128 tmp;
  v128 thresh;
  v128 total_abs;
  int off1, off2;
  off1 = OD_DIRECTION_OFFSETS_TABLE[dir][0];
  off2 = OD_DIRECTION_OFFSETS_TABLE[dir][1];
  total_abs = v128_zero();
  thresh = v128_dup_16(threshold);
  for (i = 0; i < 4; i += 2) {
    sum = v128_zero();
    row = v128_from_v64(v64_load_aligned(&in[(i + 1) * OD_FILT_BSTRIDE]),
                        v64_load_aligned(&in[i * OD_FILT_BSTRIDE]));

    /*p = in[i*OD_FILT_BSTRIDE + offset] - row*/
    tmp = v128_from_v64(v64_load_aligned(&in[(i + 1) * OD_FILT_BSTRIDE + off1]),
                        v64_load_aligned(&in[i * OD_FILT_BSTRIDE + off1]));
    p = v128_sub_16(tmp, row);
    /*if (abs(p) < thresh) sum += taps[k]*p*/
    cmp = od_cmplt_abs_epi16(p, thresh);
    p = v128_shl_n_16(p, 2);
    p = v128_and(p, cmp);
    sum = v128_add_16(sum, p);
    /*p = in[i*OD_FILT_BSTRIDE - offset] - row*/
    tmp = v128_from_v64(v64_load_aligned(&in[(i + 1) * OD_FILT_BSTRIDE - off1]),
                        v64_load_aligned(&in[i * OD_FILT_BSTRIDE - off1]));
    p = v128_sub_16(tmp, row);
    /*if (abs(p) < thresh) sum += taps[k]*p1*/
    cmp = od_cmplt_abs_epi16(p, thresh);
    p = v128_shl_n_16(p, 2);
    p = v128_and(p, cmp);
    sum = v128_add_16(sum, p);

    /*p = in[i*OD_FILT_BSTRIDE + offset] - row*/
    tmp = v128_from_v64(v64_load_aligned(&in[(i + 1) * OD_FILT_BSTRIDE + off2]),
                        v64_load_aligned(&in[i * OD_FILT_BSTRIDE + off2]));
    p = v128_sub_16(tmp, row);
    /*if (abs(p) < thresh) sum += taps[k]*p*/
    cmp = od_cmplt_abs_epi16(p, thresh);
    p = v128_and(p, cmp);
    sum = v128_add_16(sum, p);
    /*p = in[i*OD_FILT_BSTRIDE - offset] - row*/
    tmp = v128_from_v64(v64_load_aligned(&in[(i + 1) * OD_FILT_BSTRIDE - off2]),
                        v64_load_aligned(&in[i * OD_FILT_BSTRIDE - off2]));
    p = v128_sub_16(tmp, row);
    /*if (abs(p) < thresh) sum += taps[k]*p1*/
    cmp = od_cmplt_abs_epi16(p, thresh);
    p = v128_and(p, cmp);
    sum = v128_add_16(sum, p);

    /*res = row + ((sum + 8) >> 4)*/
    res = v128_add_16(sum, v128_dup_16(8));
    res = v128_shr_n_s16(res, 4);
    total_abs = v128_add_16(total_abs, v128_abs_s16(res));
    res = v128_add_16(row, res);
    v64_store_aligned(&y[i * ystride], v128_low_v64(res));
    v64_store_aligned(&y[(i + 1) * ystride], v128_high_v64(res));
  }
  return (v128_dotp_s16(total_abs, v128_dup_16(1)) + 2) >> 2;
}

int SIMD_FUNC(od_filter_dering_direction_8x8)(uint16_t *y, int ystride,
                                              const uint16_t *in, int threshold,
                                              int dir) {
  int i;
  v128 sum;
  v128 p;
  v128 cmp;
  v128 row;
  v128 res;
  v128 thresh;
  v128 total_abs;
  int off1, off2, off3;
  off1 = OD_DIRECTION_OFFSETS_TABLE[dir][0];
  off2 = OD_DIRECTION_OFFSETS_TABLE[dir][1];
  off3 = OD_DIRECTION_OFFSETS_TABLE[dir][2];
  total_abs = v128_zero();
  thresh = v128_dup_16(threshold);
  for (i = 0; i < 8; i++) {
    sum = v128_zero();
    row = v128_load_unaligned(&in[i * OD_FILT_BSTRIDE]);

    /*p = in[i*OD_FILT_BSTRIDE + offset] - row*/
    p = v128_sub_16(v128_load_unaligned(&in[i * OD_FILT_BSTRIDE + off1]), row);
    /*if (abs(p) < thresh) sum += taps[k]*p*/
    cmp = od_cmplt_abs_epi16(p, thresh);
    p = v128_add_16(p, v128_shl_n_16(p, 1));
    p = v128_and(p, cmp);
    sum = v128_add_16(sum, p);

    /*p = in[i*OD_FILT_BSTRIDE - offset] - row*/
    p = v128_sub_16(v128_load_unaligned(&in[i * OD_FILT_BSTRIDE - off1]), row);
    /*if (abs(p) < thresh) sum += taps[k]*p1*/
    cmp = od_cmplt_abs_epi16(p, thresh);
    p = v128_add_16(p, v128_shl_n_16(p, 1));
    p = v128_and(p, cmp);
    sum = v128_add_16(sum, p);

    /*p = in[i*OD_FILT_BSTRIDE + offset] - row*/
    p = v128_sub_16(v128_load_unaligned(&in[i * OD_FILT_BSTRIDE + off2]), row);
    /*if (abs(p) < thresh) sum += taps[k]*p*/
    cmp = od_cmplt_abs_epi16(p, thresh);
    p = v128_shl_n_16(p, 1);
    p = v128_and(p, cmp);
    sum = v128_add_16(sum, p);

    /*p = in[i*OD_FILT_BSTRIDE - offset] - row*/
    p = v128_sub_16(v128_load_unaligned(&in[i * OD_FILT_BSTRIDE - off2]), row);
    /*if (abs(p) < thresh) sum += taps[k]*p1*/
    cmp = od_cmplt_abs_epi16(p, thresh);
    p = v128_shl_n_16(p, 1);
    p = v128_and(p, cmp);
    sum = v128_add_16(sum, p);

    /*p = in[i*OD_FILT_BSTRIDE + offset] - row*/
    p = v128_sub_16(v128_load_unaligned(&in[i * OD_FILT_BSTRIDE + off3]), row);
    /*if (abs(p) < thresh) sum += taps[k]*p*/
    cmp = od_cmplt_abs_epi16(p, thresh);
    p = v128_and(p, cmp);
    sum = v128_add_16(sum, p);

    /*p = in[i*OD_FILT_BSTRIDE - offset] - row*/
    p = v128_sub_16(v128_load_unaligned(&in[i * OD_FILT_BSTRIDE - off3]), row);
    /*if (abs(p) < thresh) sum += taps[k]*p1*/
    cmp = od_cmplt_abs_epi16(p, thresh);
    p = v128_and(p, cmp);
    sum = v128_add_16(sum, p);

    /*res = row + ((sum + 8) >> 4)*/
    res = v128_add_16(sum, v128_dup_16(8));
    res = v128_shr_n_s16(res, 4);
    total_abs = v128_add_16(total_abs, v128_abs_s16(res));
    res = v128_add_16(row, res);
    v128_store_unaligned(&y[i * ystride], res);
  }
  return (v128_dotp_s16(total_abs, v128_dup_16(1)) + 8) >> 4;
}