FFmpeg coverage

Directory: ../../../ffmpeg/
File: src/tests/checkasm/vp8dsp.c
Date: 2026-04-23 21:43:00
Exec Total Coverage
Lines: 314 320 98.1%
Functions: 13 13 100.0%
Branches: 209 270 77.4%
Line Branch Exec Source
1 /*
2 * Copyright (c) 2016 Martin Storsjo
3 *
4 * This file is part of FFmpeg.
5 *
6 * FFmpeg is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License as published by
8 * the Free Software Foundation; either version 2 of the License, or
9 * (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
14 * GNU General Public License for more details.
15 *
16 * You should have received a copy of the GNU General Public License along
17 * with FFmpeg; if not, write to the Free Software Foundation, Inc.,
18 * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
19 */
20
21 #include <stdbool.h>
22 #include <string.h>
23
24 #include "config_components.h"
25 #include "libavcodec/vp8dsp.h"
26
27 #include "libavutil/common.h"
28 #include "libavutil/intreadwrite.h"
29 #include "libavutil/mem_internal.h"
30
31 #include "checkasm.h"
32
33 #define PIXEL_STRIDE 16
34
35 #define randomize_buffers(src, dst, stride, coef) \
36 do { \
37 int x, y; \
38 for (y = 0; y < 4; y++) { \
39 AV_WN32A((src) + y * (stride), rnd()); \
40 AV_WN32A((dst) + y * (stride), rnd()); \
41 for (x = 0; x < 4; x++) \
42 (coef)[y * 4 + x] = (src)[y * (stride) + x] - \
43 (dst)[y * (stride) + x]; \
44 } \
45 } while (0)
46
47 500 static void dct4x4(int16_t *coef)
48 {
49 int i;
50
2/2
✓ Branch 0 taken 2000 times.
✓ Branch 1 taken 500 times.
 2500 for (i = 0; i < 4; i++) {
51 2000 const int a1 = (coef[i*4 + 0] + coef[i*4 + 3]) * 8;
52 2000 const int b1 = (coef[i*4 + 1] + coef[i*4 + 2]) * 8;
53 2000 const int c1 = (coef[i*4 + 1] - coef[i*4 + 2]) * 8;
54 2000 const int d1 = (coef[i*4 + 0] - coef[i*4 + 3]) * 8;
55 2000 coef[i*4 + 0] = a1 + b1;
56 2000 coef[i*4 + 1] = (c1 * 2217 + d1 * 5352 + 14500) >> 12;
57 2000 coef[i*4 + 2] = a1 - b1;
58 2000 coef[i*4 + 3] = (d1 * 2217 - c1 * 5352 + 7500) >> 12;
59 }
60
2/2
✓ Branch 0 taken 2000 times.
✓ Branch 1 taken 500 times.
 2500 for (i = 0; i < 4; i++) {
61 2000 const int a1 = coef[i + 0*4] + coef[i + 3*4];
62 2000 const int b1 = coef[i + 1*4] + coef[i + 2*4];
63 2000 const int c1 = coef[i + 1*4] - coef[i + 2*4];
64 2000 const int d1 = coef[i + 0*4] - coef[i + 3*4];
65 2000 coef[i + 0*4] = (a1 + b1 + 7) >> 4;
66 2000 coef[i + 1*4] = ((c1 * 2217 + d1 * 5352 + 12000) >> 16) + !!d1;
67 2000 coef[i + 2*4] = (a1 - b1 + 7) >> 4;
68 2000 coef[i + 3*4] = (d1 * 2217 - c1 * 5352 + 51000) >> 16;
69 }
70 500 }
71
72 28 static void wht4x4(int16_t *coef)
73 {
74 int i;
75
2/2
✓ Branch 0 taken 112 times.
✓ Branch 1 taken 28 times.
 140 for (i = 0; i < 4; i++) {
76 112 int a1 = coef[0 * 4 + i];
77 112 int b1 = coef[1 * 4 + i];
78 112 int c1 = coef[2 * 4 + i];
79 112 int d1 = coef[3 * 4 + i];
80 int e1;
81 112 a1 += b1;
82 112 d1 -= c1;
83 112 e1 = (a1 - d1) >> 1;
84 112 b1 = e1 - b1;
85 112 c1 = e1 - c1;
86 112 a1 -= c1;
87 112 d1 += b1;
88 112 coef[0 * 4 + i] = a1;
89 112 coef[1 * 4 + i] = c1;
90 112 coef[2 * 4 + i] = d1;
91 112 coef[3 * 4 + i] = b1;
92 }
93
2/2
✓ Branch 0 taken 112 times.
✓ Branch 1 taken 28 times.
 140 for (i = 0; i < 4; i++) {
94 112 int a1 = coef[i * 4 + 0];
95 112 int b1 = coef[i * 4 + 1];
96 112 int c1 = coef[i * 4 + 2];
97 112 int d1 = coef[i * 4 + 3];
98 int e1;
99 112 a1 += b1;
100 112 d1 -= c1;
101 112 e1 = (a1 - d1) >> 1;
102 112 b1 = e1 - b1;
103 112 c1 = e1 - c1;
104 112 a1 -= c1;
105 112 d1 += b1;
106 112 coef[i * 4 + 0] = a1 * 2;
107 112 coef[i * 4 + 1] = c1 * 2;
108 112 coef[i * 4 + 2] = d1 * 2;
109 112 coef[i * 4 + 3] = b1 * 2;
110 }
111 28 }
112
113 28 static void check_idct(VP8DSPContext *d, bool is_vp7)
114 {
115 28 LOCAL_ALIGNED_16(uint8_t, src, [4 * 4]);
116 28 LOCAL_ALIGNED_16(uint8_t, dst, [4 * 4]);
117 28 LOCAL_ALIGNED_16(uint8_t, dst0, [4 * 4]);
118 28 LOCAL_ALIGNED_16(uint8_t, dst1, [4 * 4]);
119 28 LOCAL_ALIGNED_16(int16_t, coef, [4 * 4]);
120 28 LOCAL_ALIGNED_16(int16_t, subcoef0, [4 * 4]);
121 28 LOCAL_ALIGNED_16(int16_t, subcoef1, [4 * 4]);
122 int dc;
123
2/2
✓ Branch 1 taken 26 times.
✓ Branch 2 taken 2 times.
 28 declare_func_emms(AV_CPU_FLAG_MMX, void, uint8_t *dst, int16_t *block, ptrdiff_t stride);
124
125
4/4
✓ Branch 2 taken 448 times.
✓ Branch 3 taken 112 times.
✓ Branch 4 taken 112 times.
✓ Branch 5 taken 28 times.
 588 randomize_buffers(src, dst, 4, coef);
126
127 28 dct4x4(coef);
128
129
2/2
✓ Branch 0 taken 56 times.
✓ Branch 1 taken 28 times.
 84 for (dc = 0; dc <= 1; dc++) {
130
2/2
✓ Branch 0 taken 28 times.
✓ Branch 1 taken 28 times.
 56 void (*idct)(uint8_t *, int16_t *, ptrdiff_t) = dc ? d->vp8_idct_dc_add : d->vp8_idct_add;
131
132
4/4
✓ Branch 2 taken 28 times.
✓ Branch 3 taken 28 times.
✓ Branch 5 taken 7 times.
✓ Branch 6 taken 49 times.
 56 if (check_func(idct, "vp%d_idct_%sadd", 8 - is_vp7, dc ? "dc_" : "")) {
133
2/2
✓ Branch 0 taken 4 times.
✓ Branch 1 taken 3 times.
 7 if (dc) {
134 4 memset(subcoef0, 0, 4 * 4 * sizeof(int16_t));
135 4 subcoef0[0] = coef[0];
136 } else {
137 3 memcpy(subcoef0, coef, 4 * 4 * sizeof(int16_t));
138 }
139 7 memcpy(dst0, dst, 4 * 4);
140 7 memcpy(dst1, dst, 4 * 4);
141 7 memcpy(subcoef1, subcoef0, 4 * 4 * sizeof(int16_t));
142 // Note, this uses a pixel stride of 4, even though the real decoder uses a stride as a
143 // multiple of 16. If optimizations want to take advantage of that, this test needs to be
144 // updated to make it more like the h264dsp tests.
145 7 call_ref(dst0, subcoef0, 4);
146 7 call_new(dst1, subcoef1, 4);
147
1/2
✓ Branch 0 taken 7 times.
✗ Branch 1 not taken.
 7 if (memcmp(dst0, dst1, 4 * 4) ||
148
1/2
✗ Branch 0 not taken.
✓ Branch 1 taken 7 times.
 7 memcmp(subcoef0, subcoef1, 4 * 4 * sizeof(int16_t)))
149 ✗ fail();
150
151
1/8
✗ Branch 1 not taken.
✓ Branch 2 taken 7 times.
✗ Branch 39 not taken.
✗ Branch 40 not taken.
✗ Branch 41 not taken.
✗ Branch 42 not taken.
✗ Branch 43 not taken.
✗ Branch 44 not taken.
 7 bench_new(dst1, subcoef1, 4);
152 }
153 }
154 28 }
155
156 28 static void check_idct_dc4(VP8DSPContext *d, bool is_vp7)
157 {
158 28 LOCAL_ALIGNED_16(uint8_t, src, [4 * 4 * 4]);
159 28 LOCAL_ALIGNED_16(uint8_t, dst, [4 * 4 * 4]);
160 28 LOCAL_ALIGNED_16(uint8_t, dst0, [4 * 4 * 4]);
161 28 LOCAL_ALIGNED_16(uint8_t, dst1, [4 * 4 * 4]);
162 28 LOCAL_ALIGNED_16(int16_t, coef, [4], [4 * 4]);
163 28 LOCAL_ALIGNED_16(int16_t, subcoef0, [4], [4 * 4]);
164 28 LOCAL_ALIGNED_16(int16_t, subcoef1, [4], [4 * 4]);
165 int i, chroma;
166
2/2
✓ Branch 1 taken 26 times.
✓ Branch 2 taken 2 times.
 28 declare_func_emms(AV_CPU_FLAG_MMX, void, uint8_t *dst, int16_t block[4][16], ptrdiff_t stride);
167
168
2/2
✓ Branch 0 taken 56 times.
✓ Branch 1 taken 28 times.
 84 for (chroma = 0; chroma <= 1; chroma++) {
169
2/2
✓ Branch 0 taken 28 times.
✓ Branch 1 taken 28 times.
 56 void (*idct4dc)(uint8_t *, int16_t[4][16], ptrdiff_t) = chroma ? d->vp8_idct_dc_add4uv : d->vp8_idct_dc_add4y;
170
4/4
✓ Branch 2 taken 28 times.
✓ Branch 3 taken 28 times.
✓ Branch 5 taken 6 times.
✓ Branch 6 taken 50 times.
 56 if (check_func(idct4dc, "vp%d_idct_dc_add4%s", 8 - is_vp7, chroma ? "uv" : "y")) {
171
2/2
✓ Branch 0 taken 3 times.
✓ Branch 1 taken 3 times.
 6 ptrdiff_t stride = chroma ? 8 : 16;
172
2/2
✓ Branch 0 taken 3 times.
✓ Branch 1 taken 3 times.
 6 int w = chroma ? 2 : 4;
173
2/2
✓ Branch 0 taken 24 times.
✓ Branch 1 taken 6 times.
 30 for (i = 0; i < 4; i++) {
174 24 int blockx = 4 * (i % w);
175 24 int blocky = 4 * (i / w);
176
4/4
✓ Branch 2 taken 384 times.
✓ Branch 3 taken 96 times.
✓ Branch 4 taken 96 times.
✓ Branch 5 taken 24 times.
 504 randomize_buffers(src + stride * blocky + blockx, dst + stride * blocky + blockx, stride, coef[i]);
177 24 dct4x4(coef[i]);
178 24 memset(&coef[i][1], 0, 15 * sizeof(int16_t));
179 }
180
181 6 memcpy(dst0, dst, 4 * 4 * 4);
182 6 memcpy(dst1, dst, 4 * 4 * 4);
183 6 memcpy(subcoef0, coef, 4 * 4 * 4 * sizeof(int16_t));
184 6 memcpy(subcoef1, coef, 4 * 4 * 4 * sizeof(int16_t));
185 6 call_ref(dst0, subcoef0, stride);
186 6 call_new(dst1, subcoef1, stride);
187
1/2
✓ Branch 0 taken 6 times.
✗ Branch 1 not taken.
 6 if (memcmp(dst0, dst1, 4 * 4 * 4) ||
188
1/2
✗ Branch 0 not taken.
✓ Branch 1 taken 6 times.
 6 memcmp(subcoef0, subcoef1, 4 * 4 * 4 * sizeof(int16_t)))
189 ✗ fail();
190
1/8
✗ Branch 1 not taken.
✓ Branch 2 taken 6 times.
✗ Branch 39 not taken.
✗ Branch 40 not taken.
✗ Branch 41 not taken.
✗ Branch 42 not taken.
✗ Branch 43 not taken.
✗ Branch 44 not taken.
 6 bench_new(dst1, subcoef1, stride);
191 }
192 }
193
194 28 }
195
196 28 static void check_luma_dc_wht(VP8DSPContext *d, bool is_vp7)
197 {
198 28 LOCAL_ALIGNED_16(int16_t, dc, [4 * 4]);
199 28 LOCAL_ALIGNED_16(int16_t, dc0, [4 * 4]);
200 28 LOCAL_ALIGNED_16(int16_t, dc1, [4 * 4]);
201 int16_t block[4][4][16];
202 28 LOCAL_ALIGNED_16(int16_t, block0, [4], [4][16]);
203 28 LOCAL_ALIGNED_16(int16_t, block1, [4], [4][16]);
204 int dc_only;
205 int blockx, blocky;
206
2/2
✓ Branch 1 taken 26 times.
✓ Branch 2 taken 2 times.
 28 declare_func_emms(AV_CPU_FLAG_MMX, void, int16_t block[4][4][16], int16_t dc[16]);
207
208
2/2
✓ Branch 0 taken 112 times.
✓ Branch 1 taken 28 times.
 140 for (blocky = 0; blocky < 4; blocky++) {
209
2/2
✓ Branch 0 taken 448 times.
✓ Branch 1 taken 112 times.
 560 for (blockx = 0; blockx < 4; blockx++) {
210 uint8_t src[16], dst[16];
211
4/4
✓ Branch 2 taken 7168 times.
✓ Branch 3 taken 1792 times.
✓ Branch 4 taken 1792 times.
✓ Branch 5 taken 448 times.
 9408 randomize_buffers(src, dst, 4, block[blocky][blockx]);
212
213 448 dct4x4(block[blocky][blockx]);
214 448 dc[blocky * 4 + blockx] = block[blocky][blockx][0];
215 448 block[blocky][blockx][0] = rnd();
216 }
217 }
218 28 wht4x4(dc);
219
220
2/2
✓ Branch 0 taken 56 times.
✓ Branch 1 taken 28 times.
 84 for (dc_only = 0; dc_only <= 1; dc_only++) {
221
2/2
✓ Branch 0 taken 28 times.
✓ Branch 1 taken 28 times.
 56 void (*idct)(int16_t [4][4][16], int16_t [16]) = dc_only ? d->vp8_luma_dc_wht_dc : d->vp8_luma_dc_wht;
222
223
4/4
✓ Branch 2 taken 28 times.
✓ Branch 3 taken 28 times.
✓ Branch 5 taken 5 times.
✓ Branch 6 taken 51 times.
 56 if (check_func(idct, "vp%d_luma_dc_wht%s", 8 - is_vp7, dc_only ? "_dc" : "")) {
224
2/2
✓ Branch 0 taken 2 times.
✓ Branch 1 taken 3 times.
 5 if (dc_only) {
225 2 memset(dc0, 0, 16 * sizeof(int16_t));
226 2 dc0[0] = dc[0];
227 } else {
228 3 memcpy(dc0, dc, 16 * sizeof(int16_t));
229 }
230 5 memcpy(dc1, dc0, 16 * sizeof(int16_t));
231 5 memcpy(block0, block, 4 * 4 * 16 * sizeof(int16_t));
232 5 memcpy(block1, block, 4 * 4 * 16 * sizeof(int16_t));
233 5 call_ref(block0, dc0);
234 5 call_new(block1, dc1);
235
1/2
✓ Branch 0 taken 5 times.
✗ Branch 1 not taken.
 5 if (memcmp(block0, block1, 4 * 4 * 16 * sizeof(int16_t)) ||
236
1/2
✗ Branch 0 not taken.
✓ Branch 1 taken 5 times.
 5 memcmp(dc0, dc1, 16 * sizeof(int16_t)))
237 ✗ fail();
238
1/8
✗ Branch 1 not taken.
✓ Branch 2 taken 5 times.
✗ Branch 39 not taken.
✗ Branch 40 not taken.
✗ Branch 41 not taken.
✗ Branch 42 not taken.
✗ Branch 43 not taken.
✗ Branch 44 not taken.
 5 bench_new(block1, dc1);
239 }
240 }
241 28 }
242
243 #define SRC_BUF_STRIDE 32
244 #define SRC_BUF_SIZE (((size << (size < 16)) + 5) * SRC_BUF_STRIDE)
245 // The mc subpixel interpolation filter needs the 2 previous pixels in either
246 // direction, the +1 is to make sure the actual load addresses always are
247 // unaligned.
248 #define src (buf + 2 * SRC_BUF_STRIDE + 2 + 1)
249
250 #undef randomize_buffers
251 #define randomize_buffers() \
252 do { \
253 int k; \
254 for (k = 0; k < SRC_BUF_SIZE; k += 4) { \
255 AV_WN32A(buf + k, rnd()); \
256 } \
257 } while (0)
258
259 14 static void check_mc(VP8DSPContext *d)
260 {
261 14 LOCAL_ALIGNED_16(uint8_t, buf, [32 * 32]);
262 14 BUF_RECT(uint8_t, dst0, 16, 16);
263 14 BUF_RECT(uint8_t, dst1, 16, 16);
264 int type, k, dx, dy;
265
2/2
✓ Branch 1 taken 13 times.
✓ Branch 2 taken 1 times.
 14 declare_func_emms(AV_CPU_FLAG_MMX, void, uint8_t *, ptrdiff_t,
266 const uint8_t *, ptrdiff_t, int, int, int);
267
268
2/2
✓ Branch 0 taken 28 times.
✓ Branch 1 taken 14 times.
 42 for (type = 0; type < 2; type++) {
269
2/2
✓ Branch 0 taken 196 times.
✓ Branch 1 taken 28 times.
 224 for (k = 1; k < 8; k++) {
270 196 int hsize = k / 3;
271 196 int size = 16 >> hsize;
272 196 int height = (size << 1) >> (k % 3);
273
2/2
✓ Branch 0 taken 588 times.
✓ Branch 1 taken 196 times.
 784 for (dy = 0; dy < 3; dy++) {
274
2/2
✓ Branch 0 taken 1764 times.
✓ Branch 1 taken 588 times.
 2352 for (dx = 0; dx < 3; dx++) {
275 char str[100];
276
2/2
✓ Branch 0 taken 882 times.
✓ Branch 1 taken 882 times.
 1764 vp8_mc_func func = (type ? d->put_vp8_bilinear_pixels_tab : d->put_vp8_epel_pixels_tab)[hsize][dy][dx];
277
278
4/4
✓ Branch 0 taken 588 times.
✓ Branch 1 taken 1176 times.
✓ Branch 2 taken 392 times.
✓ Branch 3 taken 196 times.
 1764 if (dx || dy) {
279
2/2
✓ Branch 0 taken 784 times.
✓ Branch 1 taken 784 times.
 1568 if (type == 0) {
280 static const char *dx_names[] = { "", "h4", "h6" };
281 static const char *dy_names[] = { "", "v4", "v6" };
282 784 snprintf(str, sizeof(str), "epel%d_%s%s", size, dx_names[dx], dy_names[dy]);
283 } else {
284
4/4
✓ Branch 0 taken 588 times.
✓ Branch 1 taken 196 times.
✓ Branch 2 taken 588 times.
✓ Branch 3 taken 196 times.
 784 snprintf(str, sizeof(str), "bilin%d_%s%s", size, dx ? "h" : "", dy ? "v" : "");
285 }
286 } else {
287 196 snprintf(str, sizeof(str), "pixels%d", size);
288 }
289
290
2/2
✓ Branch 3 taken 78 times.
✓ Branch 4 taken 1686 times.
 1764 if (check_func(func, "vp8_put_%s", str)) {
291 int mx, my;
292 int i;
293
2/2
✓ Branch 0 taken 54 times.
✓ Branch 1 taken 24 times.
 78 if (type == 0) {
294
4/4
✓ Branch 0 taken 21 times.
✓ Branch 1 taken 33 times.
✓ Branch 3 taken 15 times.
✓ Branch 4 taken 18 times.
 54 mx = dx == 2 ? 2 + 2 * (rnd() % 3) : dx == 1 ? 1 + 2 * (rnd() % 4) : 0;
295
4/4
✓ Branch 0 taken 21 times.
✓ Branch 1 taken 33 times.
✓ Branch 3 taken 15 times.
✓ Branch 4 taken 18 times.
 54 my = dy == 2 ? 2 + 2 * (rnd() % 3) : dy == 1 ? 1 + 2 * (rnd() % 4) : 0;
296 } else {
297
2/2
✓ Branch 0 taken 16 times.
✓ Branch 1 taken 8 times.
 24 mx = dx ? 1 + (rnd() % 7) : 0;
298
2/2
✓ Branch 0 taken 16 times.
✓ Branch 1 taken 8 times.
 24 my = dy ? 1 + (rnd() % 7) : 0;
299 }
300
2/2
✓ Branch 1 taken 11632 times.
✓ Branch 2 taken 78 times.
 11710 randomize_buffers();
301
2/2
✓ Branch 0 taken 468 times.
✓ Branch 1 taken 78 times.
 546 for (i = -2; i <= 3; i++) {
302
4/4
✓ Branch 0 taken 390 times.
✓ Branch 1 taken 78 times.
✓ Branch 2 taken 78 times.
✓ Branch 3 taken 312 times.
 468 int val = (i == -1 || i == 2) ? 0 : 0xff;
303 // Set pixels in the first row and column to the maximum pattern,
304 // to test for potential overflows in the filter.
305 468 src[i ] = val;
306 468 src[i * SRC_BUF_STRIDE] = val;
307 }
308 78 CLEAR_BUF_RECT(dst0);
309 78 CLEAR_BUF_RECT(dst1);
310 78 call_ref(dst0, dst0_stride, src, SRC_BUF_STRIDE, height, mx, my);
311 78 call_new(dst1, dst1_stride, src, SRC_BUF_STRIDE, height, mx, my);
312 78 checkasm_check_padded(uint8_t, dst0, dst0_stride, dst1, dst1_stride, size, height, "dst");
313
1/8
✗ Branch 1 not taken.
✓ Branch 2 taken 78 times.
✗ Branch 39 not taken.
✗ Branch 40 not taken.
✗ Branch 41 not taken.
✗ Branch 42 not taken.
✗ Branch 43 not taken.
✗ Branch 44 not taken.
 78 bench_new(dst1, size, src, SRC_BUF_STRIDE, height, mx, my);
314 }
315 }
316 }
317 }
318 }
319 14 }
320
321 #undef randomize_buffers
322
323 #define setpx(a, b, c) buf[(a) + (b) * jstride] = av_clip_uint8(c)
324 // Set the pixel to c +/- [0,d]
325 #define setdx(a, b, c, d) setpx(a, b, c - (d) + (rnd() % ((d) * 2 + 1)))
326 // Set the pixel to c +/- [d,d+e] (making sure it won't be clipped)
327 #define setdx2(a, b, o, c, d, e) setpx(a, b, o = c + ((d) + (rnd() % (e))) * (c >= 128 ? -1 : 1))
328
329 290 static void randomize_loopfilter_buffers(int lineoff, int str,
330 int dir, int flim_E, int flim_I,
331 int hev_thresh, uint8_t *buf,
332 int force_hev)
333 {
334 290 uint32_t mask = 0xff;
335
2/2
✓ Branch 0 taken 154 times.
✓ Branch 1 taken 136 times.
 290 int off = dir ? lineoff : lineoff * str;
336
2/2
✓ Branch 0 taken 154 times.
✓ Branch 1 taken 136 times.
 290 int istride = dir ? 1 : str;
337
2/2
✓ Branch 0 taken 136 times.
✓ Branch 1 taken 154 times.
 290 int jstride = dir ? str : 1;
338 int i;
339
2/2
✓ Branch 0 taken 1160 times.
✓ Branch 1 taken 290 times.
 1450 for (i = 0; i < 8; i += 2) {
340 // Row 0 will trigger hev for q0/q1, row 2 will trigger hev for p0/p1,
341 // rows 4 and 6 will not trigger hev.
342 // force_hev 1 will make sure all rows trigger hev, while force_hev -1
343 // makes none of them trigger it.
344 1160 int idx = off + i * istride, p2, p1, p0, q0, q1, q2;
345 1160 setpx(idx, 0, q0 = rnd() & mask);
346
6/6
✓ Branch 0 taken 290 times.
✓ Branch 1 taken 870 times.
✓ Branch 2 taken 86 times.
✓ Branch 3 taken 204 times.
✓ Branch 4 taken 204 times.
✓ Branch 5 taken 752 times.
 1160 if (i == 0 && force_hev >= 0 || force_hev > 0)
347
2/2
✓ Branch 1 taken 196 times.
✓ Branch 2 taken 212 times.
 408 setdx2(idx, 1, q1, q0, hev_thresh + 1, flim_I - hev_thresh - 1);
348 else
349 752 setdx(idx, 1, q1 = q0, hev_thresh);
350 1160 setdx(idx, 2, q2 = q1, flim_I);
351 1160 setdx(idx, 3, q2, flim_I);
352 1160 setdx(idx, -1, p0 = q0, flim_E >> 2);
353
6/6
✓ Branch 0 taken 290 times.
✓ Branch 1 taken 870 times.
✓ Branch 2 taken 86 times.
✓ Branch 3 taken 204 times.
✓ Branch 4 taken 204 times.
✓ Branch 5 taken 752 times.
 1160 if (i == 2 && force_hev >= 0 || force_hev > 0)
354
2/2
✓ Branch 1 taken 201 times.
✓ Branch 2 taken 207 times.
 408 setdx2(idx, -2, p1, p0, hev_thresh + 1, flim_I - hev_thresh - 1);
355 else
356 752 setdx(idx, -2, p1 = p0, hev_thresh);
357 1160 setdx(idx, -3, p2 = p1, flim_I);
358 1160 setdx(idx, -4, p2, flim_I);
359 }
360 290 }
361
362 // Fill the buffer with random pixels
363 213 static void fill_loopfilter_buffers(uint8_t *buf, ptrdiff_t stride, int w, int h)
364 {
365 int x, y;
366
2/2
✓ Branch 0 taken 3408 times.
✓ Branch 1 taken 213 times.
 3621 for (y = 0; y < h; y++)
367
2/2
✓ Branch 0 taken 54528 times.
✓ Branch 1 taken 3408 times.
 57936 for (x = 0; x < w; x++)
368 54528 buf[y * stride + x] = rnd() & 0xff;
369 213 }
370
371 #define randomize_buffers(buf, lineoff, str, force_hev) \
372 randomize_loopfilter_buffers(lineoff, str, dir, flim_E, flim_I, hev_thresh, buf, force_hev)
373
374 28 static void check_loopfilter_16y(VP8DSPContext *d, bool is_vp7)
375 {
376 28 LOCAL_ALIGNED_16(uint8_t, base0, [32 + 16 * 16]);
377 28 LOCAL_ALIGNED_16(uint8_t, base1, [32 + 16 * 16]);
378 int dir, edge, force_hev;
379 28 int flim_E = 20, flim_I = 10, hev_thresh = 7;
380 28 declare_func(void, uint8_t *, ptrdiff_t, int, int, int);
381
382
2/2
✓ Branch 0 taken 56 times.
✓ Branch 1 taken 28 times.
 84 for (dir = 0; dir < 2; dir++) {
383
2/2
✓ Branch 0 taken 28 times.
✓ Branch 1 taken 28 times.
 56 int midoff = dir ? 4 * 16 : 4;
384
2/2
✓ Branch 0 taken 28 times.
✓ Branch 1 taken 28 times.
 56 int midoff_aligned = dir ? 4 * 16 : 16;
385 56 uint8_t *buf0 = base0 + midoff_aligned;
386 56 uint8_t *buf1 = base1 + midoff_aligned;
387
2/2
✓ Branch 0 taken 112 times.
✓ Branch 1 taken 56 times.
 168 for (edge = 0; edge < 2; edge++) {
388 112 void (*func)(uint8_t *, ptrdiff_t, int, int, int) = NULL;
389
4/5
✓ Branch 0 taken 28 times.
✓ Branch 1 taken 28 times.
✓ Branch 2 taken 28 times.
✓ Branch 3 taken 28 times.
✗ Branch 4 not taken.
 112 switch (dir << 1 | edge) {
390 28 case (0 << 1) | 0: func = d->vp8_h_loop_filter16y; break;
391 28 case (1 << 1) | 0: func = d->vp8_v_loop_filter16y; break;
392 28 case (0 << 1) | 1: func = d->vp8_h_loop_filter16y_inner; break;
393 28 case (1 << 1) | 1: func = d->vp8_v_loop_filter16y_inner; break;
394 }
395
6/6
✓ Branch 2 taken 56 times.
✓ Branch 3 taken 56 times.
✓ Branch 4 taken 56 times.
✓ Branch 5 taken 56 times.
✓ Branch 7 taken 17 times.
✓ Branch 8 taken 95 times.
 112 if (check_func(func, "vp%d_loop_filter16y%s_%s", 8 - is_vp7, edge ? "_inner" : "", dir ? "v" : "h")) {
396
2/2
✓ Branch 0 taken 51 times.
✓ Branch 1 taken 17 times.
 68 for (force_hev = -1; force_hev <= 1; force_hev++) {
397 51 fill_loopfilter_buffers(buf0 - midoff, 16, 16, 16);
398 51 randomize_buffers(buf0, 0, 16, force_hev);
399 51 randomize_buffers(buf0, 8, 16, force_hev);
400 51 memcpy(buf1 - midoff, buf0 - midoff, 16 * 16);
401 51 call_ref(buf0, 16, flim_E, flim_I, hev_thresh);
402 51 call_new(buf1, 16, flim_E, flim_I, hev_thresh);
403
1/2
✗ Branch 0 not taken.
✓ Branch 1 taken 51 times.
 51 if (memcmp(buf0 - midoff, buf1 - midoff, 16 * 16))
404 ✗ fail();
405 }
406 17 fill_loopfilter_buffers(buf0 - midoff, 16, 16, 16);
407 17 randomize_buffers(buf0, 0, 16, 0);
408 17 randomize_buffers(buf0, 8, 16, 0);
409
1/8
✗ Branch 1 not taken.
✓ Branch 2 taken 17 times.
✗ Branch 39 not taken.
✗ Branch 40 not taken.
✗ Branch 41 not taken.
✗ Branch 42 not taken.
✗ Branch 43 not taken.
✗ Branch 44 not taken.
 17 bench_new(buf0, 16, flim_E, flim_I, hev_thresh);
410 }
411 }
412 }
413 28 }
414
415 28 static void check_loopfilter_8uv(VP8DSPContext *d, bool is_vp7)
416 {
417 28 LOCAL_ALIGNED_16(uint8_t, base0u, [32 + 16 * 16]);
418 28 LOCAL_ALIGNED_16(uint8_t, base0v, [32 + 16 * 16]);
419 28 LOCAL_ALIGNED_16(uint8_t, base1u, [32 + 16 * 16]);
420 28 LOCAL_ALIGNED_16(uint8_t, base1v, [32 + 16 * 16]);
421 int dir, edge, force_hev;
422 28 int flim_E = 20, flim_I = 10, hev_thresh = 7;
423 28 declare_func(void, uint8_t *, uint8_t *, ptrdiff_t, int, int, int);
424
425
2/2
✓ Branch 0 taken 56 times.
✓ Branch 1 taken 28 times.
 84 for (dir = 0; dir < 2; dir++) {
426
2/2
✓ Branch 0 taken 28 times.
✓ Branch 1 taken 28 times.
 56 int midoff = dir ? 4 * 16 : 4;
427
2/2
✓ Branch 0 taken 28 times.
✓ Branch 1 taken 28 times.
 56 int midoff_aligned = dir ? 4 * 16 : 16;
428 56 uint8_t *buf0u = base0u + midoff_aligned;
429 56 uint8_t *buf0v = base0v + midoff_aligned;
430 56 uint8_t *buf1u = base1u + midoff_aligned;
431 56 uint8_t *buf1v = base1v + midoff_aligned;
432
2/2
✓ Branch 0 taken 112 times.
✓ Branch 1 taken 56 times.
 168 for (edge = 0; edge < 2; edge++) {
433 112 void (*func)(uint8_t *, uint8_t *, ptrdiff_t, int, int, int) = NULL;
434
4/5
✓ Branch 0 taken 28 times.
✓ Branch 1 taken 28 times.
✓ Branch 2 taken 28 times.
✓ Branch 3 taken 28 times.
✗ Branch 4 not taken.
 112 switch (dir << 1 | edge) {
435 28 case (0 << 1) | 0: func = d->vp8_h_loop_filter8uv; break;
436 28 case (1 << 1) | 0: func = d->vp8_v_loop_filter8uv; break;
437 28 case (0 << 1) | 1: func = d->vp8_h_loop_filter8uv_inner; break;
438 28 case (1 << 1) | 1: func = d->vp8_v_loop_filter8uv_inner; break;
439 }
440
6/6
✓ Branch 2 taken 56 times.
✓ Branch 3 taken 56 times.
✓ Branch 4 taken 56 times.
✓ Branch 5 taken 56 times.
✓ Branch 7 taken 17 times.
✓ Branch 8 taken 95 times.
 112 if (check_func(func, "vp%d_loop_filter8uv%s_%s", 8 - is_vp7, edge ? "_inner" : "", dir ? "v" : "h")) {
441
2/2
✓ Branch 0 taken 51 times.
✓ Branch 1 taken 17 times.
 68 for (force_hev = -1; force_hev <= 1; force_hev++) {
442 51 fill_loopfilter_buffers(buf0u - midoff, 16, 16, 16);
443 51 fill_loopfilter_buffers(buf0v - midoff, 16, 16, 16);
444 51 randomize_buffers(buf0u, 0, 16, force_hev);
445 51 randomize_buffers(buf0v, 0, 16, force_hev);
446 51 memcpy(buf1u - midoff, buf0u - midoff, 16 * 16);
447 51 memcpy(buf1v - midoff, buf0v - midoff, 16 * 16);
448
449 51 call_ref(buf0u, buf0v, 16, flim_E, flim_I, hev_thresh);
450 51 call_new(buf1u, buf1v, 16, flim_E, flim_I, hev_thresh);
451
1/2
✓ Branch 0 taken 51 times.
✗ Branch 1 not taken.
 51 if (memcmp(buf0u - midoff, buf1u - midoff, 16 * 16) ||
452
1/2
✗ Branch 0 not taken.
✓ Branch 1 taken 51 times.
 51 memcmp(buf0v - midoff, buf1v - midoff, 16 * 16))
453 ✗ fail();
454 }
455 17 fill_loopfilter_buffers(buf0u - midoff, 16, 16, 16);
456 17 fill_loopfilter_buffers(buf0v - midoff, 16, 16, 16);
457 17 randomize_buffers(buf0u, 0, 16, 0);
458 17 randomize_buffers(buf0v, 0, 16, 0);
459
1/8
✗ Branch 1 not taken.
✓ Branch 2 taken 17 times.
✗ Branch 39 not taken.
✗ Branch 40 not taken.
✗ Branch 41 not taken.
✗ Branch 42 not taken.
✗ Branch 43 not taken.
✗ Branch 44 not taken.
 17 bench_new(buf0u, buf0v, 16, flim_E, flim_I, hev_thresh);
460 }
461 }
462 }
463 28 }
464
465 28 static void check_loopfilter_simple(VP8DSPContext *d, bool is_vp7)
466 {
467 28 LOCAL_ALIGNED_16(uint8_t, base0, [32 + 16 * 16]);
468 28 LOCAL_ALIGNED_16(uint8_t, base1, [32 + 16 * 16]);
469 int dir;
470 28 int flim_E = 20, flim_I = 30, hev_thresh = 0;
471 28 declare_func(void, uint8_t *, ptrdiff_t, int);
472
473
2/2
✓ Branch 0 taken 56 times.
✓ Branch 1 taken 28 times.
 84 for (dir = 0; dir < 2; dir++) {
474
2/2
✓ Branch 0 taken 28 times.
✓ Branch 1 taken 28 times.
 56 int midoff = dir ? 4 * 16 : 4;
475
2/2
✓ Branch 0 taken 28 times.
✓ Branch 1 taken 28 times.
 56 int midoff_aligned = dir ? 4 * 16 : 16;
476 56 uint8_t *buf0 = base0 + midoff_aligned;
477 56 uint8_t *buf1 = base1 + midoff_aligned;
478
2/2
✓ Branch 0 taken 28 times.
✓ Branch 1 taken 28 times.
 56 void (*func)(uint8_t *, ptrdiff_t, int) = dir ? d->vp8_v_loop_filter_simple : d->vp8_h_loop_filter_simple;
479
4/4
✓ Branch 2 taken 28 times.
✓ Branch 3 taken 28 times.
✓ Branch 5 taken 9 times.
✓ Branch 6 taken 47 times.
 56 if (check_func(func, "vp%d_loop_filter_simple_%s", 8 - is_vp7, dir ? "v" : "h")) {
480 9 fill_loopfilter_buffers(buf0 - midoff, 16, 16, 16);
481 9 randomize_buffers(buf0, 0, 16, -1);
482 9 randomize_buffers(buf0, 8, 16, -1);
483 9 memcpy(buf1 - midoff, buf0 - midoff, 16 * 16);
484 9 call_ref(buf0, 16, flim_E);
485 9 call_new(buf1, 16, flim_E);
486
1/2
✗ Branch 0 not taken.
✓ Branch 1 taken 9 times.
 9 if (memcmp(buf0 - midoff, buf1 - midoff, 16 * 16))
487 ✗ fail();
488
1/8
✗ Branch 1 not taken.
✓ Branch 2 taken 9 times.
✗ Branch 39 not taken.
✗ Branch 40 not taken.
✗ Branch 41 not taken.
✗ Branch 42 not taken.
✗ Branch 43 not taken.
✗ Branch 44 not taken.
 9 bench_new(buf0, 16, flim_E);
489 }
490 }
491 28 }
492
493 28 static void checkasm_check_vp78dsp(VP8DSPContext *d, bool is_vp7)
494 {
495 #if CONFIG_VP7_DECODER
496
2/2
✓ Branch 0 taken 14 times.
✓ Branch 1 taken 14 times.
 28 if (is_vp7)
497 14 ff_vp7dsp_init(d);
498 else
499 #endif
500 14 ff_vp8dsp_init(d);
501 28 check_idct(d, is_vp7);
502 28 check_idct_dc4(d, is_vp7);
503 28 check_luma_dc_wht(d, is_vp7);
504 28 report("idct");
505 28 check_loopfilter_16y(d, is_vp7);
506 28 check_loopfilter_8uv(d, is_vp7);
507 28 check_loopfilter_simple(d, is_vp7);
508 28 report("loopfilter");
509 28 }
510
511 14 void checkasm_check_vp8dsp(void)
512 {
513 // Needs to be zeroed because not all size 16 epel functions exist.
514 14 VP8DSPContext d = { 0 };
515
516 14 ff_vp78dsp_init(&d);
517 14 check_mc(&d);
518 14 report("mc");
519 14 checkasm_check_vp78dsp(&d, false);
520 #if CONFIG_VP7_DECODER
521 14 checkasm_check_vp78dsp(&d, true);
522 #endif
523 14 }
524