GCC Code Coverage Report
Directory: ../../../ffmpeg/ Exec Total Coverage
File: src/libavcodec/hevcpred_template.c Lines: 305 311 98.1 %
Date: 2020-09-28 00:47:38 Branches: 287 303 94.7 %

Line Branch Exec Source
1
/*
2
 * HEVC video decoder
3
 *
4
 * Copyright (C) 2012 - 2013 Guillaume Martres
5
 *
6
 * This file is part of FFmpeg.
7
 *
8
 * FFmpeg is free software; you can redistribute it and/or
9
 * modify it under the terms of the GNU Lesser General Public
10
 * License as published by the Free Software Foundation; either
11
 * version 2.1 of the License, or (at your option) any later version.
12
 *
13
 * FFmpeg is distributed in the hope that it will be useful,
14
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
15
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
16
 * Lesser General Public License for more details.
17
 *
18
 * You should have received a copy of the GNU Lesser General Public
19
 * License along with FFmpeg; if not, write to the Free Software
20
 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
21
 */
22
23
#include "libavutil/pixdesc.h"
24
25
#include "bit_depth_template.c"
26
#include "hevcpred.h"
27
28
#define POS(x, y) src[(x) + stride * (y)]
29
30
40284946
static av_always_inline void FUNC(intra_pred)(HEVCContext *s, int x0, int y0,
31
                                              int log2_size, int c_idx)
32
{
33
#define PU(x) \
34
    ((x) >> s->ps.sps->log2_min_pu_size)
35
#define MVF(x, y) \
36
    (s->ref->tab_mvf[(x) + (y) * min_pu_width])
37
#define MVF_PU(x, y) \
38
    MVF(PU(x0 + ((x) * (1 << hshift))), PU(y0 + ((y) * (1 << vshift))))
39
#define IS_INTRA(x, y) \
40
    (MVF_PU(x, y).pred_flag == PF_INTRA)
41
#define MIN_TB_ADDR_ZS(x, y) \
42
    s->ps.pps->min_tb_addr_zs[(y) * (s->ps.sps->tb_mask+2) + (x)]
43
#define EXTEND(ptr, val, len)         \
44
do {                                  \
45
    pixel4 pix = PIXEL_SPLAT_X4(val); \
46
    for (i = 0; i < (len); i += 4)    \
47
        AV_WN4P(ptr + i, pix);        \
48
} while (0)
49
50
#define EXTEND_RIGHT_CIP(ptr, start, length)                                   \
51
        for (i = start; i < (start) + (length); i += 4)                        \
52
            if (!IS_INTRA(i, -1))                                              \
53
                AV_WN4P(&ptr[i], a);                                           \
54
            else                                                               \
55
                a = PIXEL_SPLAT_X4(ptr[i+3])
56
#define EXTEND_LEFT_CIP(ptr, start, length) \
57
        for (i = start; i > (start) - (length); i--) \
58
            if (!IS_INTRA(i - 1, -1)) \
59
                ptr[i - 1] = ptr[i]
60
#define EXTEND_UP_CIP(ptr, start, length)                                      \
61
        for (i = (start); i > (start) - (length); i -= 4)                      \
62
            if (!IS_INTRA(-1, i - 3))                                          \
63
                AV_WN4P(&ptr[i - 3], a);                                       \
64
            else                                                               \
65
                a = PIXEL_SPLAT_X4(ptr[i - 3])
66
#define EXTEND_DOWN_CIP(ptr, start, length)                                    \
67
        for (i = start; i < (start) + (length); i += 4)                        \
68
            if (!IS_INTRA(-1, i))                                              \
69
                AV_WN4P(&ptr[i], a);                                           \
70
            else                                                               \
71
                a = PIXEL_SPLAT_X4(ptr[i + 3])
72
73
40284946
    HEVCLocalContext *lc = s->HEVClc;
74
    int i;
75
40284946
    int hshift = s->ps.sps->hshift[c_idx];
76
40284946
    int vshift = s->ps.sps->vshift[c_idx];
77
40284946
    int size = (1 << log2_size);
78
40284946
    int size_in_luma_h = size << hshift;
79
40284946
    int size_in_tbs_h  = size_in_luma_h >> s->ps.sps->log2_min_tb_size;
80
40284946
    int size_in_luma_v = size << vshift;
81
40284946
    int size_in_tbs_v  = size_in_luma_v >> s->ps.sps->log2_min_tb_size;
82
40284946
    int x = x0 >> hshift;
83
40284946
    int y = y0 >> vshift;
84
40284946
    int x_tb = (x0 >> s->ps.sps->log2_min_tb_size) & s->ps.sps->tb_mask;
85
40284946
    int y_tb = (y0 >> s->ps.sps->log2_min_tb_size) & s->ps.sps->tb_mask;
86
87
40284946
    int cur_tb_addr = MIN_TB_ADDR_ZS(x_tb, y_tb);
88
89
40284946
    ptrdiff_t stride = s->frame->linesize[c_idx] / sizeof(pixel);
90
40284946
    pixel *src = (pixel*)s->frame->data[c_idx] + x + y * stride;
91
92
40284946
    int min_pu_width = s->ps.sps->min_pu_width;
93
94
40284946
    enum IntraPredMode mode = c_idx ? lc->tu.intra_pred_mode_c :
95
18924842
                              lc->tu.intra_pred_mode;
96
    pixel4 a;
97
    pixel  left_array[2 * MAX_TB_SIZE + 1];
98
    pixel  filtered_left_array[2 * MAX_TB_SIZE + 1];
99
    pixel  top_array[2 * MAX_TB_SIZE + 1];
100
    pixel  filtered_top_array[2 * MAX_TB_SIZE + 1];
101
102
40284946
    pixel  *left          = left_array + 1;
103
40284946
    pixel  *top           = top_array  + 1;
104
40284946
    pixel  *filtered_left = filtered_left_array + 1;
105
40284946
    pixel  *filtered_top  = filtered_top_array  + 1;
106

40284946
    int cand_bottom_left = lc->na.cand_bottom_left && cur_tb_addr > MIN_TB_ADDR_ZS( x_tb - 1, (y_tb + size_in_tbs_v) & s->ps.sps->tb_mask);
107
40284946
    int cand_left        = lc->na.cand_left;
108
40284946
    int cand_up_left     = lc->na.cand_up_left;
109
40284946
    int cand_up          = lc->na.cand_up;
110

40284946
    int cand_up_right    = lc->na.cand_up_right    && cur_tb_addr > MIN_TB_ADDR_ZS((x_tb + size_in_tbs_h) & s->ps.sps->tb_mask, y_tb - 1);
111
112
40284946
    int bottom_left_size = (FFMIN(y0 + 2 * size_in_luma_v, s->ps.sps->height) -
113
40284946
                           (y0 + size_in_luma_v)) >> vshift;
114
40284946
    int top_right_size   = (FFMIN(x0 + 2 * size_in_luma_h, s->ps.sps->width) -
115
40284946
                           (x0 + size_in_luma_h)) >> hshift;
116
117
40284946
    if (s->ps.pps->constrained_intra_pred_flag == 1) {
118
271050
        int size_in_luma_pu_v = PU(size_in_luma_v);
119
271050
        int size_in_luma_pu_h = PU(size_in_luma_h);
120
271050
        int on_pu_edge_x    = !av_mod_uintp2(x0, s->ps.sps->log2_min_pu_size);
121
271050
        int on_pu_edge_y    = !av_mod_uintp2(y0, s->ps.sps->log2_min_pu_size);
122
271050
        if (!size_in_luma_pu_h)
123
            size_in_luma_pu_h++;
124

271050
        if (cand_bottom_left == 1 && on_pu_edge_x) {
125
84160
            int x_left_pu   = PU(x0 - 1);
126
84160
            int y_bottom_pu = PU(y0 + size_in_luma_v);
127
84160
            int max = FFMIN(size_in_luma_pu_v, s->ps.sps->min_pu_height - y_bottom_pu);
128
84160
            cand_bottom_left = 0;
129
179804
            for (i = 0; i < max; i += 2)
130
95644
                cand_bottom_left |= (MVF(x_left_pu, y_bottom_pu + i).pred_flag == PF_INTRA);
131
        }
132

271050
        if (cand_left == 1 && on_pu_edge_x) {
133
266214
            int x_left_pu   = PU(x0 - 1);
134
266214
            int y_left_pu   = PU(y0);
135
266214
            int max = FFMIN(size_in_luma_pu_v, s->ps.sps->min_pu_height - y_left_pu);
136
266214
            cand_left = 0;
137
574872
            for (i = 0; i < max; i += 2)
138
308658
                cand_left |= (MVF(x_left_pu, y_left_pu + i).pred_flag == PF_INTRA);
139
        }
140
271050
        if (cand_up_left == 1) {
141
258168
            int x_left_pu   = PU(x0 - 1);
142
258168
            int y_top_pu    = PU(y0 - 1);
143
258168
            cand_up_left = MVF(x_left_pu, y_top_pu).pred_flag == PF_INTRA;
144
        }
145

271050
        if (cand_up == 1 && on_pu_edge_y) {
146
262806
            int x_top_pu    = PU(x0);
147
262806
            int y_top_pu    = PU(y0 - 1);
148
262806
            int max = FFMIN(size_in_luma_pu_h, s->ps.sps->min_pu_width - x_top_pu);
149
262806
            cand_up = 0;
150
565704
            for (i = 0; i < max; i += 2)
151
302898
                cand_up |= (MVF(x_top_pu + i, y_top_pu).pred_flag == PF_INTRA);
152
        }
153

271050
        if (cand_up_right == 1 && on_pu_edge_y) {
154
174386
            int y_top_pu    = PU(y0 - 1);
155
174386
            int x_right_pu  = PU(x0 + size_in_luma_h);
156
174386
            int max = FFMIN(size_in_luma_pu_h, s->ps.sps->min_pu_width - x_right_pu);
157
174386
            cand_up_right = 0;
158
375976
            for (i = 0; i < max; i += 2)
159
201590
                cand_up_right |= (MVF(x_right_pu + i, y_top_pu).pred_flag == PF_INTRA);
160
        }
161
271050
        memset(left, 128, 2 * MAX_TB_SIZE*sizeof(pixel));
162
271050
        memset(top , 128, 2 * MAX_TB_SIZE*sizeof(pixel));
163
271050
        top[-1] = 128;
164
    }
165
40284946
    if (cand_up_left) {
166
39093398
        left[-1] = POS(-1, -1);
167
39093398
        top[-1]  = left[-1];
168
    }
169
40284946
    if (cand_up)
170
39546160
        memcpy(top, src - stride, size * sizeof(pixel));
171
40284946
    if (cand_up_right) {
172
25816214
        memcpy(top + size, src - stride + size, size * sizeof(pixel));
173
25869522
        EXTEND(top + size + top_right_size, POS(size + top_right_size - 1, -1),
174
               size - top_right_size);
175
    }
176
40284946
    if (cand_left)
177
278022026
        for (i = 0; i < size; i++)
178
238213648
            left[i] = POS(-1, i);
179
40284946
    if (cand_bottom_left) {
180
89098348
        for (i = size; i < size + bottom_left_size; i++)
181
75872672
            left[i] = POS(-1, i);
182
13290328
        EXTEND(left + size + bottom_left_size, POS(-1, size + bottom_left_size - 1),
183
               size - bottom_left_size);
184
    }
185
186
40284946
    if (s->ps.pps->constrained_intra_pred_flag == 1) {
187


271050
        if (cand_bottom_left || cand_left || cand_up_left || cand_up || cand_up_right) {
188
535404
            int size_max_x = x0 + ((2 * size) << hshift) < s->ps.sps->width ?
189
267702
                                    2 * size : (s->ps.sps->width - x0) >> hshift;
190
535404
            int size_max_y = y0 + ((2 * size) << vshift) < s->ps.sps->height ?
191
267702
                                    2 * size : (s->ps.sps->height - y0) >> vshift;
192
267702
            int j = size + (cand_bottom_left? bottom_left_size: 0) -1;
193
267702
            if (!cand_up_right) {
194
104394
                size_max_x = x0 + ((size) << hshift) < s->ps.sps->width ?
195
104394
                                                    size : (s->ps.sps->width - x0) >> hshift;
196
            }
197
267702
            if (!cand_bottom_left) {
198
189324
                size_max_y = y0 + (( size) << vshift) < s->ps.sps->height ?
199
189324
                                                     size : (s->ps.sps->height - y0) >> vshift;
200
            }
201

267702
            if (cand_bottom_left || cand_left || cand_up_left) {
202

273404
                while (j > -1 && !IS_INTRA(-1, j))
203
17280
                    j--;
204
256124
                if (!IS_INTRA(-1, j)) {
205
                    j = 0;
206
                    while (j < size_max_x && !IS_INTRA(j, -1))
207
                        j++;
208
                    EXTEND_LEFT_CIP(top, j, j + 1);
209
                    left[-1] = top[-1];
210
                }
211
            } else {
212
11578
                j = 0;
213

18858
                while (j < size_max_x && !IS_INTRA(j, -1))
214
7280
                    j++;
215
11578
                if (j > 0)
216
818
                    if (x0 > 0) {
217

8208
                        EXTEND_LEFT_CIP(top, j, j + 1);
218
                    } else {
219

650
                        EXTEND_LEFT_CIP(top, j, j);
220
58
                        top[-1] = top[0];
221
                    }
222
11578
                left[-1] = top[-1];
223
            }
224
267702
            left[-1] = top[-1];
225

267702
            if (cand_bottom_left || cand_left) {
226
253796
                a = PIXEL_SPLAT_X4(left[-1]);
227

702646
                EXTEND_DOWN_CIP(left, 0, size_max_y);
228
            }
229
267702
            if (!cand_left)
230
38910
                EXTEND(left, left[-1], size);
231
267702
            if (!cand_bottom_left)
232
453010
                EXTEND(left + size, left[size - 1], size);
233

267702
            if (x0 != 0 && y0 != 0) {
234
260370
                a = PIXEL_SPLAT_X4(left[size_max_y - 1]);
235

714714
                EXTEND_UP_CIP(left, size_max_y - 1, size_max_y);
236
260370
                if (!IS_INTRA(-1, - 1))
237
14852
                    left[-1] = left[0];
238
7332
            } else if (x0 == 0) {
239
8996
                EXTEND(left, 0, size_max_y);
240
            } else {
241
4228
                a = PIXEL_SPLAT_X4(left[size_max_y - 1]);
242

16018
                EXTEND_UP_CIP(left, size_max_y - 1, size_max_y);
243
            }
244
267702
            top[-1] = left[-1];
245
267702
            if (y0 != 0) {
246
263474
                a = PIXEL_SPLAT_X4(left[-1]);
247

844462
                EXTEND_RIGHT_CIP(top, 0, size_max_x);
248
            }
249
        }
250
    }
251
    // Infer the unavailable samples
252
40284946
    if (!cand_bottom_left) {
253
27059270
        if (cand_left) {
254
67105122
            EXTEND(left + size, left[size - 1], size);
255
475872
        } else if (cand_up_left) {
256
9640
            EXTEND(left, left[-1], 2 * size);
257
2328
            cand_left = 1;
258
473544
        } else if (cand_up) {
259
452766
            left[-1] = top[0];
260
2530562
            EXTEND(left, left[-1], 2 * size);
261
452766
            cand_up_left = 1;
262
452766
            cand_left    = 1;
263
20778
        } else if (cand_up_right) {
264
2132
            EXTEND(top, top[size], size);
265
680
            left[-1] = top[size];
266
3584
            EXTEND(left, left[-1], 2 * size);
267
680
            cand_up      = 1;
268
680
            cand_up_left = 1;
269
680
            cand_left    = 1;
270
        } else { // No samples available
271
20098
            left[-1] = (1 << (BIT_DEPTH - 1));
272
132910
            EXTEND(top,  left[-1], 2 * size);
273
132910
            EXTEND(left, left[-1], 2 * size);
274
        }
275
    }
276
277
40284946
    if (!cand_left)
278
78332
        EXTEND(left, left[size], size);
279
40284946
    if (!cand_up_left) {
280
738102
        left[-1] = left[0];
281
    }
282
40284946
    if (!cand_up)
283
2265358
        EXTEND(top, left[-1], size);
284
40284946
    if (!cand_up_right)
285
35979736
        EXTEND(top + size, top[size - 1], size);
286
287
40284946
    top[-1] = left[-1];
288
289
    // Filtering process
290

40284946
    if (!s->ps.sps->intra_smoothing_disabled_flag && (c_idx == 0  || s->ps.sps->chroma_format_idc == 3)) {
291

19376732
        if (mode != INTRA_DC && size != 4){
292
6185182
            int intra_hor_ver_dist_thresh[] = { 7, 1, 0 };
293

6185182
            int min_dist_vert_hor = FFMIN(FFABS((int)(mode - 26U)),
294
                                          FFABS((int)(mode - 10U)));
295
6185182
            if (min_dist_vert_hor > intra_hor_ver_dist_thresh[log2_size - 3]) {
296
2484896
                int threshold = 1 << (BIT_DEPTH - 5);
297

2484896
                if (s->ps.sps->sps_strong_intra_smoothing_enable_flag && c_idx == 0 &&
298
280734
                    log2_size == 5 &&
299
280734
                    FFABS(top[-1]  + top[63]  - 2 * top[31])  < threshold &&
300
107688
                    FFABS(left[-1] + left[63] - 2 * left[31]) < threshold) {
301
                    // We can't just overwrite values in top because it could be
302
                    // a pointer into src
303
58292
                    filtered_top[-1] = top[-1];
304
58292
                    filtered_top[63] = top[63];
305
3730688
                    for (i = 0; i < 63; i++)
306
3672396
                        filtered_top[i] = ((64 - (i + 1)) * top[-1] +
307
3672396
                                           (i + 1)  * top[63] + 32) >> 6;
308
3730688
                    for (i = 0; i < 63; i++)
309
3672396
                        left[i] = ((64 - (i + 1)) * left[-1] +
310
3672396
                                   (i + 1)  * left[63] + 32) >> 6;
311
58292
                    top = filtered_top;
312
                } else {
313
2426604
                    filtered_left[2 * size - 1] = left[2 * size - 1];
314
2426604
                    filtered_top[2 * size - 1]  = top[2 * size - 1];
315
68308576
                    for (i = 2 * size - 2; i >= 0; i--)
316
65881972
                        filtered_left[i] = (left[i + 1] + 2 * left[i] +
317
65881972
                                            left[i - 1] + 2) >> 2;
318
2426604
                    filtered_top[-1]  =
319
2426604
                    filtered_left[-1] = (left[0] + 2 * left[-1] + top[0] + 2) >> 2;
320
68308576
                    for (i = 2 * size - 2; i >= 0; i--)
321
65881972
                        filtered_top[i] = (top[i + 1] + 2 * top[i] +
322
65881972
                                           top[i - 1] + 2) >> 2;
323
2426604
                    left = filtered_left;
324
2426604
                    top  = filtered_top;
325
                }
326
            }
327
        }
328
    }
329
330
40284946
    switch (mode) {
331
8323380
    case INTRA_PLANAR:
332
8323380
        s->hpc.pred_planar[log2_size - 2]((uint8_t *)src, (uint8_t *)top,
333
                                          (uint8_t *)left, stride);
334
8323380
        break;
335
4998978
    case INTRA_DC:
336
4998978
        s->hpc.pred_dc((uint8_t *)src, (uint8_t *)top,
337
                       (uint8_t *)left, stride, log2_size, c_idx);
338
4998978
        break;
339
26962588
    default:
340
26962588
        s->hpc.pred_angular[log2_size - 2]((uint8_t *)src, (uint8_t *)top,
341
                                           (uint8_t *)left, stride, c_idx,
342
                                           mode);
343
26962588
        break;
344
    }
345
}
346
347
#define INTRA_PRED(size)                                                            \
348
static void FUNC(intra_pred_ ## size)(HEVCContext *s, int x0, int y0, int c_idx)    \
349
{                                                                                   \
350
    FUNC(intra_pred)(s, x0, y0, size, c_idx);                                       \
351
}
352
353
28655564
INTRA_PRED(2)
354
4171506
INTRA_PRED(3)
355
1372199
INTRA_PRED(4)
356
270986
INTRA_PRED(5)
357
358
#undef INTRA_PRED
359
360
8323380
static av_always_inline void FUNC(pred_planar)(uint8_t *_src, const uint8_t *_top,
361
                                  const uint8_t *_left, ptrdiff_t stride,
362
                                  int trafo_size)
363
{
364
    int x, y;
365
8323380
    pixel *src        = (pixel *)_src;
366
8323380
    const pixel *top  = (const pixel *)_top;
367
8323380
    const pixel *left = (const pixel *)_left;
368
8323380
    int size = 1 << trafo_size;
369
61368572
    for (y = 0; y < size; y++)
370
592239464
        for (x = 0; x < size; x++)
371
539194272
            POS(x, y) = ((size - 1 - x) * left[y] + (x + 1) * top[size]  +
372
539194272
                         (size - 1 - y) * top[x]  + (y + 1) * left[size] + size) >> (trafo_size + 1);
373
}
374
375
#define PRED_PLANAR(size)\
376
static void FUNC(pred_planar_ ## size)(uint8_t *src, const uint8_t *top,        \
377
                                       const uint8_t *left, ptrdiff_t stride)   \
378
{                                                                               \
379
    FUNC(pred_planar)(src, top, left, stride, size + 2);                        \
380
}
381
382
5686586
PRED_PLANAR(0)
383
895580
PRED_PLANAR(1)
384
346585
PRED_PLANAR(2)
385
76232
PRED_PLANAR(3)
386
387
#undef PRED_PLANAR
388
389
4998978
static void FUNC(pred_dc)(uint8_t *_src, const uint8_t *_top,
390
                          const uint8_t *_left,
391
                          ptrdiff_t stride, int log2_size, int c_idx)
392
{
393
    int i, j, x, y;
394
4998978
    int size          = (1 << log2_size);
395
4998978
    pixel *src        = (pixel *)_src;
396
4998978
    const pixel *top  = (const pixel *)_top;
397
4998978
    const pixel *left = (const pixel *)_left;
398
4998978
    int dc            = size;
399
    pixel4 a;
400
39412674
    for (i = 0; i < size; i++)
401
34413696
        dc += left[i] + top[i];
402
403
4998978
    dc >>= log2_size + 1;
404
405
4998978
    a = PIXEL_SPLAT_X4(dc);
406
407
39412674
    for (i = 0; i < size; i++)
408
130960848
        for (j = 0; j < size; j+=4)
409
96547152
            AV_WN4P(&POS(j, i), a);
410
411

4998978
    if (c_idx == 0 && size < 32) {
412
1807204
        POS(0, 0) = (left[0] + 2 * dc + top[0] + 2) >> 2;
413
11898704
        for (x = 1; x < size; x++)
414
10091500
            POS(x, 0) = (top[x] + 3 * dc + 2) >> 2;
415
11898704
        for (y = 1; y < size; y++)
416
10091500
            POS(0, y) = (left[y] + 3 * dc + 2) >> 2;
417
    }
418
}
419
420
26962588
static av_always_inline void FUNC(pred_angular)(uint8_t *_src,
421
                                                const uint8_t *_top,
422
                                                const uint8_t *_left,
423
                                                ptrdiff_t stride, int c_idx,
424
                                                int mode, int size)
425
{
426
    int x, y;
427
26962588
    pixel *src        = (pixel *)_src;
428
26962588
    const pixel *top  = (const pixel *)_top;
429
26962588
    const pixel *left = (const pixel *)_left;
430
431
    static const int intra_pred_angle[] = {
432
         32,  26,  21,  17, 13,  9,  5, 2, 0, -2, -5, -9, -13, -17, -21, -26, -32,
433
        -26, -21, -17, -13, -9, -5, -2, 0, 2,  5,  9, 13,  17,  21,  26,  32
434
    };
435
    static const int inv_angle[] = {
436
        -4096, -1638, -910, -630, -482, -390, -315, -256, -315, -390, -482,
437
        -630, -910, -1638, -4096
438
    };
439
440
26962588
    int angle = intra_pred_angle[mode - 2];
441
    pixel ref_array[3 * MAX_TB_SIZE + 4];
442
26962588
    pixel *ref_tmp = ref_array + size;
443
    const pixel *ref;
444
26962588
    int last = (size * angle) >> 5;
445
446
26962588
    if (mode >= 18) {
447
14623072
        ref = top - 1;
448

14623072
        if (angle < 0 && last < -1) {
449
12498356
            for (x = 0; x <= size; x += 4)
450
8898408
                AV_WN4P(&ref_tmp[x], AV_RN4P(&top[x - 1]));
451
16747932
            for (x = last; x <= -1; x++)
452
13147984
                ref_tmp[x] = left[-1 + ((x * inv_angle[mode - 11] + 128) >> 8)];
453
3599948
            ref = ref_tmp;
454
        }
455
456
98003072
        for (y = 0; y < size; y++) {
457
83380000
            int idx  = ((y + 1) * angle) >> 5;
458
83380000
            int fact = ((y + 1) * angle) & 31;
459
83380000
            if (fact) {
460
152284414
                for (x = 0; x < size; x += 4) {
461
102209752
                    POS(x    , y) = ((32 - fact) * ref[x + idx + 1] +
462
102209752
                                           fact  * ref[x + idx + 2] + 16) >> 5;
463
102209752
                    POS(x + 1, y) = ((32 - fact) * ref[x + 1 + idx + 1] +
464
102209752
                                           fact  * ref[x + 1 + idx + 2] + 16) >> 5;
465
102209752
                    POS(x + 2, y) = ((32 - fact) * ref[x + 2 + idx + 1] +
466
102209752
                                           fact  * ref[x + 2 + idx + 2] + 16) >> 5;
467
102209752
                    POS(x + 3, y) = ((32 - fact) * ref[x + 3 + idx + 1] +
468
102209752
                                           fact  * ref[x + 3 + idx + 2] + 16) >> 5;
469
                }
470
            } else {
471
105692050
                for (x = 0; x < size; x += 4)
472
72386712
                    AV_WN4P(&POS(x, y), AV_RN4P(&ref[x + idx + 1]));
473
            }
474
        }
475

14623072
        if (mode == 26 && c_idx == 0 && size < 32) {
476
11448492
            for (y = 0; y < size; y++)
477
9806328
                POS(0, y) = av_clip_pixel(top[0] + ((left[y] - left[-1]) >> 1));
478
        }
479
    } else {
480
12339516
        ref = left - 1;
481

12339516
        if (angle < 0 && last < -1) {
482
8462610
            for (x = 0; x <= size; x += 4)
483
6051188
                AV_WN4P(&ref_tmp[x], AV_RN4P(&left[x - 1]));
484
10578980
            for (x = last; x <= -1; x++)
485
8167558
                ref_tmp[x] = top[-1 + ((x * inv_angle[mode - 11] + 128) >> 8)];
486
2411422
            ref = ref_tmp;
487
        }
488
489
84120452
        for (x = 0; x < size; x++) {
490
71780936
            int idx  = ((x + 1) * angle) >> 5;
491
71780936
            int fact = ((x + 1) * angle) & 31;
492
71780936
            if (fact) {
493
497232654
                for (y = 0; y < size; y++) {
494
445622656
                    POS(x, y) = ((32 - fact) * ref[y + idx + 1] +
495
445622656
                                       fact  * ref[y + idx + 2] + 16) >> 5;
496
                }
497
            } else {
498
200766554
                for (y = 0; y < size; y++)
499
180595616
                    POS(x, y) = ref[y + idx + 1];
500
            }
501
        }
502

12339516
        if (mode == 10 && c_idx == 0 && size < 32) {
503
2107436
            for (x = 0; x < size; x += 4) {
504
1255212
                POS(x,     0) = av_clip_pixel(left[0] + ((top[x    ] - top[-1]) >> 1));
505
1255212
                POS(x + 1, 0) = av_clip_pixel(left[0] + ((top[x + 1] - top[-1]) >> 1));
506
1255212
                POS(x + 2, 0) = av_clip_pixel(left[0] + ((top[x + 2] - top[-1]) >> 1));
507
1255212
                POS(x + 3, 0) = av_clip_pixel(left[0] + ((top[x + 3] - top[-1]) >> 1));
508
            }
509
        }
510
    }
511
}
512
513
19779734
static void FUNC(pred_angular_0)(uint8_t *src, const uint8_t *top,
514
                                 const uint8_t *left,
515
                                 ptrdiff_t stride, int c_idx, int mode)
516
{
517
19779734
    FUNC(pred_angular)(src, top, left, stride, c_idx, mode, 1 << 2);
518
}
519
520
5394446
static void FUNC(pred_angular_1)(uint8_t *src, const uint8_t *top,
521
                                 const uint8_t *left,
522
                                 ptrdiff_t stride, int c_idx, int mode)
523
{
524
5394446
    FUNC(pred_angular)(src, top, left, stride, c_idx, mode, 1 << 3);
525
}
526
527
1521414
static void FUNC(pred_angular_2)(uint8_t *src, const uint8_t *top,
528
                                 const uint8_t *left,
529
                                 ptrdiff_t stride, int c_idx, int mode)
530
{
531
1521414
    FUNC(pred_angular)(src, top, left, stride, c_idx, mode, 1 << 4);
532
}
533
534
266994
static void FUNC(pred_angular_3)(uint8_t *src, const uint8_t *top,
535
                                 const uint8_t *left,
536
                                 ptrdiff_t stride, int c_idx, int mode)
537
{
538
266994
    FUNC(pred_angular)(src, top, left, stride, c_idx, mode, 1 << 5);
539
}
540
541
#undef EXTEND_LEFT_CIP
542
#undef EXTEND_RIGHT_CIP
543
#undef EXTEND_UP_CIP
544
#undef EXTEND_DOWN_CIP
545
#undef IS_INTRA
546
#undef MVF_PU
547
#undef MVF
548
#undef PU
549
#undef EXTEND
550
#undef MIN_TB_ADDR_ZS
551
#undef POS