GCC Code Coverage Report
Directory: ../../../ffmpeg/ Exec Total Coverage
File: src/libavcodec/vp3dsp.c Lines: 265 266 99.6 %
Date: 2020-09-25 14:59:26 Branches: 45 48 93.8 %

Line Branch Exec Source
1
/*
2
 * Copyright (C) 2004 The FFmpeg project
3
 *
4
 * This file is part of FFmpeg.
5
 *
6
 * FFmpeg is free software; you can redistribute it and/or
7
 * modify it under the terms of the GNU Lesser General Public
8
 * License as published by the Free Software Foundation; either
9
 * version 2.1 of the License, or (at your option) any later version.
10
 *
11
 * FFmpeg is distributed in the hope that it will be useful,
12
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
14
 * Lesser General Public License for more details.
15
 *
16
 * You should have received a copy of the GNU Lesser General Public
17
 * License along with FFmpeg; if not, write to the Free Software
18
 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
19
 */
20
21
/**
22
 * @file
23
 * Standard C DSP-oriented functions cribbed from the original VP3
24
 * source code.
25
 */
26
27
#include "libavutil/attributes.h"
28
#include "libavutil/common.h"
29
#include "libavutil/intreadwrite.h"
30
#include "libavutil/avassert.h"
31
32
#include "avcodec.h"
33
#include "rnd_avg.h"
34
#include "vp3dsp.h"
35
36
#define IdctAdjustBeforeShift 8
37
#define xC1S7 64277
38
#define xC2S6 60547
39
#define xC3S5 54491
40
#define xC4S4 46341
41
#define xC5S3 36410
42
#define xC6S2 25080
43
#define xC7S1 12785
44
45
#define M(a, b) ((int)((SUINT)(a) * (b)) >> 16)
46
47
1496608
static av_always_inline void idct(uint8_t *dst, ptrdiff_t stride,
48
                                  int16_t *input, int type)
49
{
50
1496608
    int16_t *ip = input;
51
52
    int A, B, C, D, Ad, Bd, Cd, Dd, E, F, G, H;
53
    int Ed, Gd, Add, Bdd, Fd, Hd;
54
55
    int i;
56
57
    /* Inverse DCT on the rows now */
58
13469472
    for (i = 0; i < 8; i++) {
59
        /* Check for non-zero values */
60
11972864
        if (ip[0 * 8] | ip[1 * 8] | ip[2 * 8] | ip[3 * 8] |
61
11972864
            ip[4 * 8] | ip[5 * 8] | ip[6 * 8] | ip[7 * 8]) {
62
2240890
            A = M(xC1S7, ip[1 * 8]) + M(xC7S1, ip[7 * 8]);
63
2240890
            B = M(xC7S1, ip[1 * 8]) - M(xC1S7, ip[7 * 8]);
64
2240890
            C = M(xC3S5, ip[3 * 8]) + M(xC5S3, ip[5 * 8]);
65
2240890
            D = M(xC3S5, ip[5 * 8]) - M(xC5S3, ip[3 * 8]);
66
67
2240890
            Ad = M(xC4S4, (A - C));
68
2240890
            Bd = M(xC4S4, (B - D));
69
70
2240890
            Cd = A + C;
71
2240890
            Dd = B + D;
72
73
2240890
            E = M(xC4S4, (ip[0 * 8] + ip[4 * 8]));
74
2240890
            F = M(xC4S4, (ip[0 * 8] - ip[4 * 8]));
75
76
2240890
            G = M(xC2S6, ip[2 * 8]) + M(xC6S2, ip[6 * 8]);
77
2240890
            H = M(xC6S2, ip[2 * 8]) - M(xC2S6, ip[6 * 8]);
78
79
2240890
            Ed = E - G;
80
2240890
            Gd = E + G;
81
82
2240890
            Add = F + Ad;
83
2240890
            Bdd = Bd - H;
84
85
2240890
            Fd = F - Ad;
86
2240890
            Hd = Bd + H;
87
88
            /*  Final sequence of operations over-write original inputs. */
89
2240890
            ip[0 * 8] = Gd + Cd;
90
2240890
            ip[7 * 8] = Gd - Cd;
91
92
2240890
            ip[1 * 8] = Add + Hd;
93
2240890
            ip[2 * 8] = Add - Hd;
94
95
2240890
            ip[3 * 8] = Ed + Dd;
96
2240890
            ip[4 * 8] = Ed - Dd;
97
98
2240890
            ip[5 * 8] = Fd + Bdd;
99
2240890
            ip[6 * 8] = Fd - Bdd;
100
        }
101
102
11972864
        ip += 1;            /* next row */
103
    }
104
105
1496608
    ip = input;
106
107
13469472
    for (i = 0; i < 8; i++) {
108
        /* Check for non-zero values (bitwise or faster than ||) */
109
11972864
        if (ip[1] | ip[2] | ip[3] |
110
11972864
            ip[4] | ip[5] | ip[6] | ip[7]) {
111
3885848
            A = M(xC1S7, ip[1]) + M(xC7S1, ip[7]);
112
3885848
            B = M(xC7S1, ip[1]) - M(xC1S7, ip[7]);
113
3885848
            C = M(xC3S5, ip[3]) + M(xC5S3, ip[5]);
114
3885848
            D = M(xC3S5, ip[5]) - M(xC5S3, ip[3]);
115
116
3885848
            Ad = M(xC4S4, (A - C));
117
3885848
            Bd = M(xC4S4, (B - D));
118
119
3885848
            Cd = A + C;
120
3885848
            Dd = B + D;
121
122
3885848
            E = M(xC4S4, (ip[0] + ip[4])) + 8;
123
3885848
            F = M(xC4S4, (ip[0] - ip[4])) + 8;
124
125
3885848
            if (type == 1) { // HACK
126
741480
                E += 16 * 128;
127
741480
                F += 16 * 128;
128
            }
129
130
3885848
            G = M(xC2S6, ip[2]) + M(xC6S2, ip[6]);
131
3885848
            H = M(xC6S2, ip[2]) - M(xC2S6, ip[6]);
132
133
3885848
            Ed = E - G;
134
3885848
            Gd = E + G;
135
136
3885848
            Add = F + Ad;
137
3885848
            Bdd = Bd - H;
138
139
3885848
            Fd = F - Ad;
140
3885848
            Hd = Bd + H;
141
142
            /* Final sequence of operations over-write original inputs. */
143
3885848
            if (type == 1) {
144
741480
                dst[0 * stride] = av_clip_uint8((Gd + Cd) >> 4);
145
741480
                dst[7 * stride] = av_clip_uint8((Gd - Cd) >> 4);
146
147
741480
                dst[1 * stride] = av_clip_uint8((Add + Hd) >> 4);
148
741480
                dst[2 * stride] = av_clip_uint8((Add - Hd) >> 4);
149
150
741480
                dst[3 * stride] = av_clip_uint8((Ed + Dd) >> 4);
151
741480
                dst[4 * stride] = av_clip_uint8((Ed - Dd) >> 4);
152
153
741480
                dst[5 * stride] = av_clip_uint8((Fd + Bdd) >> 4);
154
741480
                dst[6 * stride] = av_clip_uint8((Fd - Bdd) >> 4);
155
            } else {
156
3144368
                dst[0 * stride] = av_clip_uint8(dst[0 * stride] + ((Gd + Cd) >> 4));
157
3144368
                dst[7 * stride] = av_clip_uint8(dst[7 * stride] + ((Gd - Cd) >> 4));
158
159
3144368
                dst[1 * stride] = av_clip_uint8(dst[1 * stride] + ((Add + Hd) >> 4));
160
3144368
                dst[2 * stride] = av_clip_uint8(dst[2 * stride] + ((Add - Hd) >> 4));
161
162
3144368
                dst[3 * stride] = av_clip_uint8(dst[3 * stride] + ((Ed + Dd) >> 4));
163
3144368
                dst[4 * stride] = av_clip_uint8(dst[4 * stride] + ((Ed - Dd) >> 4));
164
165
3144368
                dst[5 * stride] = av_clip_uint8(dst[5 * stride] + ((Fd + Bdd) >> 4));
166
3144368
                dst[6 * stride] = av_clip_uint8(dst[6 * stride] + ((Fd - Bdd) >> 4));
167
            }
168
        } else {
169
8087016
            if (type == 1) {
170
2034952
                dst[0*stride] =
171
2034952
                dst[1*stride] =
172
2034952
                dst[2*stride] =
173
2034952
                dst[3*stride] =
174
2034952
                dst[4*stride] =
175
2034952
                dst[5*stride] =
176
2034952
                dst[6*stride] =
177
2034952
                dst[7*stride] = av_clip_uint8(128 + ((xC4S4 * ip[0] + (IdctAdjustBeforeShift << 16)) >> 20));
178
            } else {
179
6052064
                if (ip[0]) {
180
1482015
                    int v = (xC4S4 * ip[0] + (IdctAdjustBeforeShift << 16)) >> 20;
181
1482015
                    dst[0 * stride] = av_clip_uint8(dst[0 * stride] + v);
182
1482015
                    dst[1 * stride] = av_clip_uint8(dst[1 * stride] + v);
183
1482015
                    dst[2 * stride] = av_clip_uint8(dst[2 * stride] + v);
184
1482015
                    dst[3 * stride] = av_clip_uint8(dst[3 * stride] + v);
185
1482015
                    dst[4 * stride] = av_clip_uint8(dst[4 * stride] + v);
186
1482015
                    dst[5 * stride] = av_clip_uint8(dst[5 * stride] + v);
187
1482015
                    dst[6 * stride] = av_clip_uint8(dst[6 * stride] + v);
188
1482015
                    dst[7 * stride] = av_clip_uint8(dst[7 * stride] + v);
189
                }
190
            }
191
        }
192
193
11972864
        ip += 8;            /* next column */
194
11972864
        dst++;
195
    }
196
1496608
}
197
198
404050
static av_always_inline void idct10(uint8_t *dst, ptrdiff_t stride,
199
                                    int16_t *input, int type)
200
{
201
404050
    int16_t *ip = input;
202
203
    int A, B, C, D, Ad, Bd, Cd, Dd, E, F, G, H;
204
    int Ed, Gd, Add, Bdd, Fd, Hd;
205
206
    int i;
207
208
    /* Inverse DCT on the rows now */
209
2020250
    for (i = 0; i < 4; i++) {
210
        /* Check for non-zero values */
211
1616200
        if (ip[0 * 8] | ip[1 * 8] | ip[2 * 8] | ip[3 * 8]) {
212
110501
            A =  M(xC1S7, ip[1 * 8]);
213
110501
            B =  M(xC7S1, ip[1 * 8]);
214
110501
            C =  M(xC3S5, ip[3 * 8]);
215
110501
            D = -M(xC5S3, ip[3 * 8]);
216
217
110501
            Ad = M(xC4S4, (A - C));
218
110501
            Bd = M(xC4S4, (B - D));
219
220
110501
            Cd = A + C;
221
110501
            Dd = B + D;
222
223
110501
            E = M(xC4S4, ip[0 * 8]);
224
110501
            F = E;
225
226
110501
            G = M(xC2S6, ip[2 * 8]);
227
110501
            H = M(xC6S2, ip[2 * 8]);
228
229
110501
            Ed = E - G;
230
110501
            Gd = E + G;
231
232
110501
            Add = F + Ad;
233
110501
            Bdd = Bd - H;
234
235
110501
            Fd = F - Ad;
236
110501
            Hd = Bd + H;
237
238
            /* Final sequence of operations over-write original inputs */
239
110501
            ip[0 * 8] = Gd + Cd;
240
110501
            ip[7 * 8] = Gd - Cd;
241
242
110501
            ip[1 * 8] = Add + Hd;
243
110501
            ip[2 * 8] = Add - Hd;
244
245
110501
            ip[3 * 8] = Ed + Dd;
246
110501
            ip[4 * 8] = Ed - Dd;
247
248
110501
            ip[5 * 8] = Fd + Bdd;
249
110501
            ip[6 * 8] = Fd - Bdd;
250
251
        }
252
253
1616200
        ip += 1;
254
    }
255
256
404050
    ip = input;
257
258
3636450
    for (i = 0; i < 8; i++) {
259
        /* Check for non-zero values (bitwise or faster than ||) */
260
3232400
        if (ip[0] | ip[1] | ip[2] | ip[3]) {
261
695916
            A =  M(xC1S7, ip[1]);
262
695916
            B =  M(xC7S1, ip[1]);
263
695916
            C =  M(xC3S5, ip[3]);
264
695916
            D = -M(xC5S3, ip[3]);
265
266
695916
            Ad = M(xC4S4, (A - C));
267
695916
            Bd = M(xC4S4, (B - D));
268
269
695916
            Cd = A + C;
270
695916
            Dd = B + D;
271
272
695916
            E = M(xC4S4, ip[0]);
273
695916
            if (type == 1)
274
227082
                E += 16 * 128;
275
695916
            F = E;
276
277
695916
            G = M(xC2S6, ip[2]);
278
695916
            H = M(xC6S2, ip[2]);
279
280
695916
            Ed = E - G;
281
695916
            Gd = E + G;
282
283
695916
            Add = F + Ad;
284
695916
            Bdd = Bd - H;
285
286
695916
            Fd = F - Ad;
287
695916
            Hd = Bd + H;
288
289
695916
            Gd += 8;
290
695916
            Add += 8;
291
695916
            Ed += 8;
292
695916
            Fd += 8;
293
294
            /* Final sequence of operations over-write original inputs. */
295
695916
            if (type == 1) {
296
227082
                dst[0 * stride] = av_clip_uint8((Gd + Cd) >> 4);
297
227082
                dst[7 * stride] = av_clip_uint8((Gd - Cd) >> 4);
298
299
227082
                dst[1 * stride] = av_clip_uint8((Add + Hd) >> 4);
300
227082
                dst[2 * stride] = av_clip_uint8((Add - Hd) >> 4);
301
302
227082
                dst[3 * stride] = av_clip_uint8((Ed + Dd) >> 4);
303
227082
                dst[4 * stride] = av_clip_uint8((Ed - Dd) >> 4);
304
305
227082
                dst[5 * stride] = av_clip_uint8((Fd + Bdd) >> 4);
306
227082
                dst[6 * stride] = av_clip_uint8((Fd - Bdd) >> 4);
307
            } else {
308
468834
                dst[0 * stride] = av_clip_uint8(dst[0 * stride] + ((Gd + Cd) >> 4));
309
468834
                dst[7 * stride] = av_clip_uint8(dst[7 * stride] + ((Gd - Cd) >> 4));
310
311
468834
                dst[1 * stride] = av_clip_uint8(dst[1 * stride] + ((Add + Hd) >> 4));
312
468834
                dst[2 * stride] = av_clip_uint8(dst[2 * stride] + ((Add - Hd) >> 4));
313
314
468834
                dst[3 * stride] = av_clip_uint8(dst[3 * stride] + ((Ed + Dd) >> 4));
315
468834
                dst[4 * stride] = av_clip_uint8(dst[4 * stride] + ((Ed - Dd) >> 4));
316
317
468834
                dst[5 * stride] = av_clip_uint8(dst[5 * stride] + ((Fd + Bdd) >> 4));
318
468834
                dst[6 * stride] = av_clip_uint8(dst[6 * stride] + ((Fd - Bdd) >> 4));
319
            }
320
        } else {
321
2536484
            if (type == 1) {
322
52470
                dst[0*stride] =
323
52470
                dst[1*stride] =
324
52470
                dst[2*stride] =
325
52470
                dst[3*stride] =
326
52470
                dst[4*stride] =
327
52470
                dst[5*stride] =
328
52470
                dst[6*stride] =
329
52470
                dst[7*stride] = 128;
330
            }
331
        }
332
333
3232400
        ip += 8;
334
3232400
        dst++;
335
    }
336
404050
}
337
338
34944
void ff_vp3dsp_idct10_put(uint8_t *dest, ptrdiff_t stride, int16_t *block)
339
{
340
34944
    idct10(dest, stride, block, 1);
341
34944
    memset(block, 0, sizeof(*block) * 64);
342
34944
}
343
344
369106
void ff_vp3dsp_idct10_add(uint8_t *dest, ptrdiff_t stride, int16_t *block)
345
{
346
369106
    idct10(dest, stride, block, 2);
347
369106
    memset(block, 0, sizeof(*block) * 64);
348
369106
}
349
350
347054
static void vp3_idct_put_c(uint8_t *dest /* align 8 */, ptrdiff_t stride,
351
                           int16_t *block /* align 16 */)
352
{
353
347054
    idct(dest, stride, block, 1);
354
347054
    memset(block, 0, sizeof(*block) * 64);
355
347054
}
356
357
1149554
static void vp3_idct_add_c(uint8_t *dest /* align 8 */, ptrdiff_t stride,
358
                           int16_t *block /* align 16 */)
359
{
360
1149554
    idct(dest, stride, block, 2);
361
1149554
    memset(block, 0, sizeof(*block) * 64);
362
1149554
}
363
364
127148
static void vp3_idct_dc_add_c(uint8_t *dest /* align 8 */, ptrdiff_t stride,
365
                              int16_t *block /* align 16 */)
366
{
367
127148
    int i, dc = (block[0] + 15) >> 5;
368
369
1144332
    for (i = 0; i < 8; i++) {
370
1017184
        dest[0] = av_clip_uint8(dest[0] + dc);
371
1017184
        dest[1] = av_clip_uint8(dest[1] + dc);
372
1017184
        dest[2] = av_clip_uint8(dest[2] + dc);
373
1017184
        dest[3] = av_clip_uint8(dest[3] + dc);
374
1017184
        dest[4] = av_clip_uint8(dest[4] + dc);
375
1017184
        dest[5] = av_clip_uint8(dest[5] + dc);
376
1017184
        dest[6] = av_clip_uint8(dest[6] + dc);
377
1017184
        dest[7] = av_clip_uint8(dest[7] + dc);
378
1017184
        dest   += stride;
379
    }
380
127148
    block[0] = 0;
381
127148
}
382
383
465940
static av_always_inline void vp3_v_loop_filter_c(uint8_t *first_pixel, ptrdiff_t stride,
384
                                                 int *bounding_values, int count)
385
{
386
    unsigned char *end;
387
    int filter_value;
388
465940
    const ptrdiff_t nstride = -stride;
389
390
4514716
    for (end = first_pixel + count; first_pixel < end; first_pixel++) {
391
4048776
        filter_value = (first_pixel[2 * nstride] - first_pixel[stride]) +
392
4048776
                       (first_pixel[0] - first_pixel[nstride]) * 3;
393
4048776
        filter_value = bounding_values[(filter_value + 4) >> 3];
394
395
4048776
        first_pixel[nstride] = av_clip_uint8(first_pixel[nstride] + filter_value);
396
4048776
        first_pixel[0]       = av_clip_uint8(first_pixel[0] - filter_value);
397
    }
398
465940
}
399
400
492808
static av_always_inline void vp3_h_loop_filter_c(uint8_t *first_pixel, ptrdiff_t stride,
401
                                                 int *bounding_values, int count)
402
{
403
    unsigned char *end;
404
    int filter_value;
405
406
4800344
    for (end = first_pixel + count * stride; first_pixel != end; first_pixel += stride) {
407
4307536
        filter_value = (first_pixel[-2] - first_pixel[1]) +
408
4307536
                       (first_pixel[ 0] - first_pixel[-1]) * 3;
409
4307536
        filter_value = bounding_values[(filter_value + 4) >> 3];
410
411
4307536
        first_pixel[-1] = av_clip_uint8(first_pixel[-1] + filter_value);
412
4307536
        first_pixel[ 0] = av_clip_uint8(first_pixel[ 0] - filter_value);
413
    }
414
492808
}
415
416
#define LOOP_FILTER(prefix, suffix, dim, count) \
417
void prefix##_##dim##_loop_filter_##count##suffix(uint8_t *first_pixel, ptrdiff_t stride, \
418
                                int *bounding_values) \
419
{ \
420
    vp3_##dim##_loop_filter_c(first_pixel, stride, bounding_values, count); \
421
}
422
423
385626
static LOOP_FILTER(vp3,_c, v, 8)
424
401540
static LOOP_FILTER(vp3,_c, h, 8)
425
80314
LOOP_FILTER(ff_vp3dsp, , v, 12)
426
91268
LOOP_FILTER(ff_vp3dsp, , h, 12)
427
428
333697
static void put_no_rnd_pixels_l2(uint8_t *dst, const uint8_t *src1,
429
                                 const uint8_t *src2, ptrdiff_t stride, int h)
430
{
431
    int i;
432
433
3003273
    for (i = 0; i < h; i++) {
434
        uint32_t a, b;
435
436
2669576
        a = AV_RN32(&src1[i * stride]);
437
2669576
        b = AV_RN32(&src2[i * stride]);
438
2669576
        AV_WN32A(&dst[i * stride], no_rnd_avg32(a, b));
439
2669576
        a = AV_RN32(&src1[i * stride + 4]);
440
2669576
        b = AV_RN32(&src2[i * stride + 4]);
441
2669576
        AV_WN32A(&dst[i * stride + 4], no_rnd_avg32(a, b));
442
    }
443
333697
}
444
445
31
av_cold void ff_vp3dsp_init(VP3DSPContext *c, int flags)
446
{
447
31
    c->put_no_rnd_pixels_l2 = put_no_rnd_pixels_l2;
448
449
31
    c->idct_put      = vp3_idct_put_c;
450
31
    c->idct_add      = vp3_idct_add_c;
451
31
    c->idct_dc_add   = vp3_idct_dc_add_c;
452
31
    c->v_loop_filter = c->v_loop_filter_unaligned = vp3_v_loop_filter_8_c;
453
31
    c->h_loop_filter = c->h_loop_filter_unaligned = vp3_h_loop_filter_8_c;
454
455
    if (ARCH_ARM)
456
        ff_vp3dsp_init_arm(c, flags);
457
    if (ARCH_PPC)
458
        ff_vp3dsp_init_ppc(c, flags);
459
    if (ARCH_X86)
460
31
        ff_vp3dsp_init_x86(c, flags);
461
    if (ARCH_MIPS)
462
        ff_vp3dsp_init_mips(c, flags);
463
31
}
464
465
/*
466
 * This function initializes the loop filter boundary limits if the frame's
467
 * quality index is different from the previous frame's.
468
 *
469
 * where sizeof(bounding_values_array) is 256 * sizeof(int)
470
 *
471
 * The filter_limit_values may not be larger than 127.
472
 */
473
470
void ff_vp3dsp_set_bounding_values(int * bounding_values_array, int filter_limit)
474
{
475
470
    int *bounding_values = bounding_values_array + 127;
476
    int x;
477
    int value;
478
479
470
    av_assert0(filter_limit < 128U);
480
481
    /* set up the bounding values */
482
470
    memset(bounding_values_array, 0, 256 * sizeof(int));
483
3126
    for (x = 0; x < filter_limit; x++) {
484
2656
        bounding_values[-x] = -x;
485
2656
        bounding_values[x] = x;
486
    }
487

3126
    for (x = value = filter_limit; x < 128 && value; x++, value--) {
488
2656
        bounding_values[ x] =  value;
489
2656
        bounding_values[-x] = -value;
490
    }
491
470
    if (value)
492
        bounding_values[128] = value;
493
470
    bounding_values[129] = bounding_values[130] = filter_limit * 0x02020202;
494
470
}