FFmpeg coverage

Directory: ../../../ffmpeg/
File: src/libavcodec/x86/me_cmp_init.c
Date: 2025-11-28 06:05:02
Exec Total Coverage
Lines: 48 54 88.9%
Functions: 3 3 100.0%
Branches: 15 20 75.0%
Line Branch Exec Source
1 /*
2 * SIMD-optimized motion estimation
3 * Copyright (c) 2000, 2001 Fabrice Bellard
4 * Copyright (c) 2002-2004 Michael Niedermayer <michaelni@gmx.at>
5 *
6 * MMX optimization by Nick Kurshev <nickols_k@mail.ru>
7 *
8 * This file is part of FFmpeg.
9 *
10 * FFmpeg is free software; you can redistribute it and/or
11 * modify it under the terms of the GNU Lesser General Public
12 * License as published by the Free Software Foundation; either
13 * version 2.1 of the License, or (at your option) any later version.
14 *
15 * FFmpeg is distributed in the hope that it will be useful,
16 * but WITHOUT ANY WARRANTY; without even the implied warranty of
17 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
18 * Lesser General Public License for more details.
19 *
20 * You should have received a copy of the GNU Lesser General Public
21 * License along with FFmpeg; if not, write to the Free Software
22 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
23 */
24
25 #include "libavutil/attributes.h"
26 #include "libavutil/cpu.h"
27 #include "libavutil/x86/cpu.h"
28 #include "libavcodec/me_cmp.h"
29 #include "libavcodec/mpegvideoenc.h"
30
31 int ff_sum_abs_dctelem_sse2(const int16_t *block);
32 int ff_sum_abs_dctelem_ssse3(const int16_t *block);
33 int ff_sse8_sse2(MPVEncContext *v, const uint8_t *pix1, const uint8_t *pix2,
34 ptrdiff_t stride, int h);
35 int ff_sse16_sse2(MPVEncContext *v, const uint8_t *pix1, const uint8_t *pix2,
36 ptrdiff_t stride, int h);
37 int ff_hf_noise8_ssse3(const uint8_t *pix1, ptrdiff_t stride, int h);
38 int ff_hf_noise16_ssse3(const uint8_t *pix1, ptrdiff_t stride, int h);
39 int ff_sad8_mmxext(MPVEncContext *v, const uint8_t *pix1, const uint8_t *pix2,
40 ptrdiff_t stride, int h);
41 int ff_sad16_sse2(MPVEncContext *v, const uint8_t *pix1, const uint8_t *pix2,
42 ptrdiff_t stride, int h);
43 int ff_sad16u_sse2(MPVEncContext *v, const uint8_t *pix1, const uint8_t *pix2,
44 ptrdiff_t stride, int h);
45 int ff_sad8_x2_mmxext(MPVEncContext *v, const uint8_t *pix1, const uint8_t *pix2,
46 ptrdiff_t stride, int h);
47 int ff_sad16_x2_sse2(MPVEncContext *v, const uint8_t *pix1, const uint8_t *pix2,
48 ptrdiff_t stride, int h);
49 int ff_sad8_y2_mmxext(MPVEncContext *v, const uint8_t *pix1, const uint8_t *pix2,
50 ptrdiff_t stride, int h);
51 int ff_sad16_y2_sse2(MPVEncContext *v, const uint8_t *pix1, const uint8_t *pix2,
52 ptrdiff_t stride, int h);
53 int ff_sad8_approx_xy2_mmxext(MPVEncContext *v, const uint8_t *pix1, const uint8_t *pix2,
54 ptrdiff_t stride, int h);
55 int ff_sad8_xy2_sse2(MPVEncContext *v, const uint8_t *pix1, const uint8_t *pix2,
56 ptrdiff_t stride, int h);
57 int ff_sad16_approx_xy2_sse2(MPVEncContext *v, const uint8_t *pix1, const uint8_t *pix2,
58 ptrdiff_t stride, int h);
59 int ff_sad16_xy2_sse2(MPVEncContext *v, const uint8_t *pix1, const uint8_t *pix2,
60 ptrdiff_t stride, int h);
61 int ff_vsad_intra8_mmxext(MPVEncContext *v, const uint8_t *pix1, const uint8_t *pix2,
62 ptrdiff_t stride, int h);
63 int ff_vsad_intra16_sse2(MPVEncContext *v, const uint8_t *pix1, const uint8_t *pix2,
64 ptrdiff_t stride, int h);
65 int ff_vsad_intra16u_sse2(MPVEncContext *v, const uint8_t *pix1, const uint8_t *pix2,
66 ptrdiff_t stride, int h);
67 int ff_vsad8_approx_mmxext(MPVEncContext *v, const uint8_t *pix1, const uint8_t *pix2,
68 ptrdiff_t stride, int h);
69 int ff_vsad16_approx_sse2(MPVEncContext *v, const uint8_t *pix1, const uint8_t *pix2,
70 ptrdiff_t stride, int h);
71 int ff_vsad16u_approx_sse2(MPVEncContext *v, const uint8_t *pix1, const uint8_t *pix2,
72 ptrdiff_t stride, int h);
73
74 #define hadamard_func(cpu) \
75 int ff_hadamard8_diff_ ## cpu(MPVEncContext *s, const uint8_t *src1, \
76 const uint8_t *src2, ptrdiff_t stride, int h); \
77 int ff_hadamard8_diff16_ ## cpu(MPVEncContext *s, const uint8_t *src1, \
78 const uint8_t *src2, ptrdiff_t stride, int h);
79
80 hadamard_func(sse2)
81 hadamard_func(ssse3)
82
83 #if HAVE_X86ASM
84 16 static int nsse16_ssse3(MPVEncContext *c, const uint8_t *pix1, const uint8_t *pix2,
85 ptrdiff_t stride, int h)
86 {
87 16 int score1 = ff_sse16_sse2(c, pix1, pix2, stride, h);
88 16 int score2 = ff_hf_noise16_ssse3(pix1, stride, h) -
89 16 ff_hf_noise16_ssse3(pix2, stride, h);
90
91
1/2
✗ Branch 0 not taken.
✓ Branch 1 taken 16 times.
 16 if (c)
92 ✗ return score1 + FFABS(score2) * c->c.avctx->nsse_weight;
93 else
94 16 return score1 + FFABS(score2) * 8;
95 }
96
97 16 static int nsse8_ssse3(MPVEncContext *c, const uint8_t *pix1, const uint8_t *pix2,
98 ptrdiff_t stride, int h)
99 {
100 16 int score1 = ff_sse8_sse2(c, pix1, pix2, stride, h);
101 16 int score2 = ff_hf_noise8_ssse3(pix1, stride, h) -
102 16 ff_hf_noise8_ssse3(pix2, stride, h);
103
104
1/2
✗ Branch 0 not taken.
✓ Branch 1 taken 16 times.
 16 if (c)
105 ✗ return score1 + FFABS(score2) * c->c.avctx->nsse_weight;
106 else
107 16 return score1 + FFABS(score2) * 8;
108 }
109
110 #endif /* HAVE_X86ASM */
111
112 1069 av_cold void ff_me_cmp_init_x86(MECmpContext *c, AVCodecContext *avctx)
113 {
114 #if HAVE_X86ASM
115 1069 int cpu_flags = av_get_cpu_flags();
116
117
2/2
✓ Branch 0 taken 69 times.
✓ Branch 1 taken 1000 times.
 1069 if (EXTERNAL_MMXEXT(cpu_flags)) {
118 69 c->sad[1] = ff_sad8_mmxext;
119
120 69 c->pix_abs[1][0] = ff_sad8_mmxext;
121 69 c->pix_abs[1][1] = ff_sad8_x2_mmxext;
122 69 c->pix_abs[1][2] = ff_sad8_y2_mmxext;
123
124 69 c->vsad[5] = ff_vsad_intra8_mmxext;
125
126
2/2
✓ Branch 0 taken 56 times.
✓ Branch 1 taken 13 times.
 69 if (!(avctx->flags & AV_CODEC_FLAG_BITEXACT)) {
127 56 c->pix_abs[1][3] = ff_sad8_approx_xy2_mmxext;
128
129 56 c->vsad[1] = ff_vsad8_approx_mmxext;
130 }
131 }
132
133
2/2
✓ Branch 0 taken 67 times.
✓ Branch 1 taken 1002 times.
 1069 if (EXTERNAL_SSE2(cpu_flags)) {
134 67 c->sse[0] = ff_sse16_sse2;
135 67 c->sse[1] = ff_sse8_sse2;
136 67 c->sum_abs_dctelem = ff_sum_abs_dctelem_sse2;
137
138 67 c->pix_abs[0][0] = ff_sad16_sse2;
139 67 c->pix_abs[0][1] = ff_sad16_x2_sse2;
140 67 c->pix_abs[0][2] = ff_sad16_y2_sse2;
141 67 c->pix_abs[0][3] = ff_sad16_xy2_sse2;
142
143 67 c->hadamard8_diff[0] = ff_hadamard8_diff16_sse2;
144 67 c->hadamard8_diff[1] = ff_hadamard8_diff_sse2;
145
1/2
✓ Branch 0 taken 67 times.
✗ Branch 1 not taken.
 67 if (avctx->codec_id != AV_CODEC_ID_SNOW) {
146 67 c->sad[0] = ff_sad16_sse2;
147
148 67 c->vsad[4] = ff_vsad_intra16_sse2;
149
2/2
✓ Branch 0 taken 56 times.
✓ Branch 1 taken 11 times.
 67 if (!(avctx->flags & AV_CODEC_FLAG_BITEXACT)) {
150 56 c->vsad[0] = ff_vsad16_approx_sse2;
151 }
152 } else {
153 // Snow does not abide by the alignment requirements
154 // of blk1, so we use special versions without them for it.
155 ✗ c->sad[0] = ff_sad16u_sse2;
156
157 ✗ c->vsad[4] = ff_vsad_intra16u_sse2;
158 ✗ if (!(avctx->flags & AV_CODEC_FLAG_BITEXACT)) {
159 ✗ c->vsad[0] = ff_vsad16u_approx_sse2;
160 }
161 }
162
2/2
✓ Branch 0 taken 11 times.
✓ Branch 1 taken 56 times.
 67 if (avctx->flags & AV_CODEC_FLAG_BITEXACT) {
163 11 c->pix_abs[1][3] = ff_sad8_xy2_sse2;
164 } else {
165 56 c->pix_abs[0][3] = ff_sad16_approx_xy2_sse2;
166 }
167 }
168
169
2/2
✓ Branch 0 taken 65 times.
✓ Branch 1 taken 1004 times.
 1069 if (EXTERNAL_SSSE3(cpu_flags)) {
170 65 c->nsse[0] = nsse16_ssse3;
171 65 c->nsse[1] = nsse8_ssse3;
172
173 65 c->sum_abs_dctelem = ff_sum_abs_dctelem_ssse3;
174 65 c->hadamard8_diff[0] = ff_hadamard8_diff16_ssse3;
175 65 c->hadamard8_diff[1] = ff_hadamard8_diff_ssse3;
176 }
177 #endif
178 1069 }
179