Line | Branch | Exec | Source |
---|---|---|---|
1 | /* | ||
2 | * This file is part of FFmpeg. | ||
3 | * | ||
4 | * FFmpeg is free software; you can redistribute it and/or | ||
5 | * modify it under the terms of the GNU Lesser General Public | ||
6 | * License as published by the Free Software Foundation; either | ||
7 | * version 2.1 of the License, or (at your option) any later version. | ||
8 | * | ||
9 | * FFmpeg is distributed in the hope that it will be useful, | ||
10 | * but WITHOUT ANY WARRANTY; without even the implied warranty of | ||
11 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | ||
12 | * Lesser General Public License for more details. | ||
13 | * | ||
14 | * You should have received a copy of the GNU Lesser General Public | ||
15 | * License along with FFmpeg; if not, write to the Free Software | ||
16 | * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA | ||
17 | */ | ||
18 | |||
19 | #include "libavutil/cpu.h" | ||
20 | #include "libavutil/x86/cpu.h" | ||
21 | #include "libavfilter/scene_sad.h" | ||
22 | |||
23 | #define SCENE_SAD_FUNC(FUNC_NAME, ASM_FUNC_NAME, MMSIZE) \ | ||
24 | void ASM_FUNC_NAME(SCENE_SAD_PARAMS); \ | ||
25 | \ | ||
26 | static void FUNC_NAME(SCENE_SAD_PARAMS) { \ | ||
27 | uint64_t sad[MMSIZE / 8] = {0}; \ | ||
28 | ptrdiff_t awidth = width & ~(MMSIZE - 1); \ | ||
29 | *sum = 0; \ | ||
30 | ASM_FUNC_NAME(src1, stride1, src2, stride2, awidth, height, sad); \ | ||
31 | for (int i = 0; i < MMSIZE / 8; i++) \ | ||
32 | *sum += sad[i]; \ | ||
33 | ff_scene_sad_c(src1 + awidth, stride1, \ | ||
34 | src2 + awidth, stride2, \ | ||
35 | width - awidth, height, sad); \ | ||
36 | *sum += sad[0]; \ | ||
37 | } | ||
38 | |||
39 | #define SCENE_SAD16_FUNC(FUNC_NAME, ASM_FUNC_NAME, MMSIZE) \ | ||
40 | void ASM_FUNC_NAME(SCENE_SAD_PARAMS); \ | ||
41 | \ | ||
42 | static void FUNC_NAME(SCENE_SAD_PARAMS) { \ | ||
43 | uint64_t sad[MMSIZE / 8] = {0}; \ | ||
44 | ptrdiff_t bytes = (width << 1) & ~(MMSIZE - 1); \ | ||
45 | *sum = 0; \ | ||
46 | ASM_FUNC_NAME(src1, stride1, src2, stride2, bytes, height, sad); \ | ||
47 | for (int i = 0; i < MMSIZE / 8; i++) \ | ||
48 | *sum += sad[i]; \ | ||
49 | ff_scene_sad16_c(src1 + bytes, stride1, \ | ||
50 | src2 + bytes, stride2, \ | ||
51 | width - (bytes >> 1), height, sad); \ | ||
52 | *sum += sad[0]; \ | ||
53 | } | ||
54 | |||
55 | #if HAVE_X86ASM | ||
56 |
2/2✓ Branch 1 taken 4 times.
✓ Branch 2 taken 2 times.
|
6 | SCENE_SAD_FUNC(scene_sad_sse2, ff_scene_sad8_sse2, 16) |
57 | #if HAVE_AVX2_EXTERNAL | ||
58 |
2/2✓ Branch 1 taken 13460 times.
✓ Branch 2 taken 3365 times.
|
16825 | SCENE_SAD_FUNC(scene_sad_avx2, ff_scene_sad8_avx2, 32) |
59 |
2/2✓ Branch 1 taken 16 times.
✓ Branch 2 taken 4 times.
|
20 | SCENE_SAD16_FUNC(scene_sad16_avx2, ff_scene_sad16_avx2, 32) |
60 | #endif | ||
61 | #if HAVE_AVX512_EXTERNAL | ||
62 | ✗ | SCENE_SAD_FUNC(scene_sad_avx512, ff_scene_sad8_avx512, 64) | |
63 | ✗ | SCENE_SAD16_FUNC(scene_sad16_avx512, ff_scene_sad16_avx512, 64) | |
64 | #endif | ||
65 | #endif | ||
66 | |||
67 | 87 | ff_scene_sad_fn ff_scene_sad_get_fn_x86(int depth) | |
68 | { | ||
69 | #if HAVE_X86ASM | ||
70 | 87 | int cpu_flags = av_get_cpu_flags(); | |
71 |
2/2✓ Branch 0 taken 20 times.
✓ Branch 1 taken 67 times.
|
87 | if (depth <= 8) { |
72 | #if HAVE_AVX512_EXTERNAL | ||
73 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 20 times.
|
20 | if (EXTERNAL_AVX512(cpu_flags)) |
74 | ✗ | return scene_sad_avx512; | |
75 | #endif | ||
76 | #if HAVE_AVX2_EXTERNAL | ||
77 |
3/4✓ Branch 0 taken 4 times.
✓ Branch 1 taken 16 times.
✓ Branch 2 taken 4 times.
✗ Branch 3 not taken.
|
20 | if (EXTERNAL_AVX2_FAST(cpu_flags)) |
78 | 4 | return scene_sad_avx2; | |
79 | #endif | ||
80 |
2/2✓ Branch 0 taken 8 times.
✓ Branch 1 taken 8 times.
|
16 | if (EXTERNAL_SSE2(cpu_flags)) |
81 | 8 | return scene_sad_sse2; | |
82 |
2/2✓ Branch 0 taken 54 times.
✓ Branch 1 taken 13 times.
|
67 | } else if (depth < 16) { /* this routine is only safe up to 15 bits */ |
83 | #if HAVE_AVX512_EXTERNAL | ||
84 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 54 times.
|
54 | if (EXTERNAL_AVX512(cpu_flags)) |
85 | ✗ | return scene_sad16_avx512; | |
86 | #endif | ||
87 | #if HAVE_AVX2_EXTERNAL | ||
88 |
3/4✓ Branch 0 taken 4 times.
✓ Branch 1 taken 50 times.
✓ Branch 2 taken 4 times.
✗ Branch 3 not taken.
|
54 | if (EXTERNAL_AVX2_FAST(cpu_flags)) |
89 | 4 | return scene_sad16_avx2; | |
90 | #endif | ||
91 | } | ||
92 | #endif | ||
93 | 71 | return NULL; | |
94 | } | ||
95 |