FFmpeg coverage

Directory: ../../../ffmpeg/
File: src/libavcodec/x86/me_cmp_init.c
Date: 2026-08-15 14:54:27
Exec Total Coverage
Lines: 50 56 89.3%
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 "config.h"
26 #include "libavutil/attributes.h"
27 #include "libavutil/cpu.h"
28 #include "libavutil/x86/cpu.h"
29 #include "libavcodec/me_cmp.h"
30 #include "libavcodec/mpegvideoenc.h"
31
32 int ff_sum_abs_dctelem_sse2(const int16_t *block);
33 int ff_sum_abs_dctelem_ssse3(const int16_t *block);
34 int ff_sse8_sse2(MPVEncContext *v, const uint8_t *pix1, const uint8_t *pix2,
35 ptrdiff_t stride, int h);
36 int ff_sse16_sse2(MPVEncContext *v, const uint8_t *pix1, const uint8_t *pix2,
37 ptrdiff_t stride, int h);
38 int ff_hf_noise8_ssse3(const uint8_t *pix1, ptrdiff_t stride, int h);
39 int ff_hf_noise16_ssse3(const uint8_t *pix1, ptrdiff_t stride, int h);
40 int ff_sad8_mmxext(MPVEncContext *v, const uint8_t *pix1, const uint8_t *pix2,
41 ptrdiff_t stride, int h);
42 int ff_sad16_sse2(MPVEncContext *v, const uint8_t *pix1, const uint8_t *pix2,
43 ptrdiff_t stride, int h);
44 int ff_sad16u_sse2(MPVEncContext *v, const uint8_t *pix1, const uint8_t *pix2,
45 ptrdiff_t stride, int h);
46 int ff_sad8_x2_mmxext(MPVEncContext *v, const uint8_t *pix1, const uint8_t *pix2,
47 ptrdiff_t stride, int h);
48 int ff_sad16_x2_sse2(MPVEncContext *v, const uint8_t *pix1, const uint8_t *pix2,
49 ptrdiff_t stride, int h);
50 int ff_sad8_y2_mmxext(MPVEncContext *v, const uint8_t *pix1, const uint8_t *pix2,
51 ptrdiff_t stride, int h);
52 int ff_sad16_y2_sse2(MPVEncContext *v, const uint8_t *pix1, const uint8_t *pix2,
53 ptrdiff_t stride, int h);
54 int ff_sad8_approx_xy2_mmxext(MPVEncContext *v, const uint8_t *pix1, const uint8_t *pix2,
55 ptrdiff_t stride, int h);
56 int ff_sad8_xy2_sse2(MPVEncContext *v, const uint8_t *pix1, const uint8_t *pix2,
57 ptrdiff_t stride, int h);
58 int ff_sad16_approx_xy2_sse2(MPVEncContext *v, const uint8_t *pix1, const uint8_t *pix2,
59 ptrdiff_t stride, int h);
60 int ff_sad16_xy2_sse2(MPVEncContext *v, const uint8_t *pix1, const uint8_t *pix2,
61 ptrdiff_t stride, int h);
62 int ff_vsad_intra8_mmxext(MPVEncContext *v, const uint8_t *pix1, const uint8_t *pix2,
63 ptrdiff_t stride, int h);
64 int ff_vsad_intra16_sse2(MPVEncContext *v, const uint8_t *pix1, const uint8_t *pix2,
65 ptrdiff_t stride, int h);
66 int ff_vsad_intra16u_sse2(MPVEncContext *v, const uint8_t *pix1, const uint8_t *pix2,
67 ptrdiff_t stride, int h);
68 int ff_vsad8_approx_mmxext(MPVEncContext *v, const uint8_t *pix1, const uint8_t *pix2,
69 ptrdiff_t stride, int h);
70 int ff_vsad16_approx_sse2(MPVEncContext *v, const uint8_t *pix1, const uint8_t *pix2,
71 ptrdiff_t stride, int h);
72 int ff_vsad16u_approx_sse2(MPVEncContext *v, const uint8_t *pix1, const uint8_t *pix2,
73 ptrdiff_t stride, int h);
74 int ff_median_sad16_ssse3(MPVEncContext *v, const uint8_t *pix1, const uint8_t *pix2,
75 ptrdiff_t stride, int h);
76 int ff_median_sad8_ssse3(MPVEncContext *v, const uint8_t *pix1, const uint8_t *pix2,
77 ptrdiff_t stride, int h);
78
79 #define hadamard_func(cpu) \
80 int ff_hadamard8_diff_ ## cpu(MPVEncContext *s, const uint8_t *src1, \
81 const uint8_t *src2, ptrdiff_t stride, int h); \
82 int ff_hadamard8_diff16_ ## cpu(MPVEncContext *s, const uint8_t *src1, \
83 const uint8_t *src2, ptrdiff_t stride, int h);
84
85 hadamard_func(sse2)
86 hadamard_func(ssse3)
87
88 17 static int nsse16_ssse3(MPVEncContext *c, const uint8_t *pix1, const uint8_t *pix2,
89 ptrdiff_t stride, int h)
90 {
91 17 int score1 = ff_sse16_sse2(c, pix1, pix2, stride, h);
92 17 int score2 = ff_hf_noise16_ssse3(pix1, stride, h) -
93 17 ff_hf_noise16_ssse3(pix2, stride, h);
94
95
1/2
✗ Branch 0 not taken.
✓ Branch 1 taken 17 times.
 17 if (c)
96 ✗ return score1 + FFABS(score2) * c->c.avctx->nsse_weight;
97 else
98 17 return score1 + FFABS(score2) * 8;
99 }
100
101 17 static int nsse8_ssse3(MPVEncContext *c, const uint8_t *pix1, const uint8_t *pix2,
102 ptrdiff_t stride, int h)
103 {
104 17 int score1 = ff_sse8_sse2(c, pix1, pix2, stride, h);
105 17 int score2 = ff_hf_noise8_ssse3(pix1, stride, h) -
106 17 ff_hf_noise8_ssse3(pix2, stride, h);
107
108
1/2
✗ Branch 0 not taken.
✓ Branch 1 taken 17 times.
 17 if (c)
109 ✗ return score1 + FFABS(score2) * c->c.avctx->nsse_weight;
110 else
111 17 return score1 + FFABS(score2) * 8;
112 }
113
114 1583 av_cold void ff_me_cmp_init_x86(MECmpContext *c, AVCodecContext *avctx)
115 {
116 1583 int cpu_flags = av_get_cpu_flags();
117
118
2/2
✓ Branch 0 taken 253 times.
✓ Branch 1 taken 1330 times.
 1583 if (EXTERNAL_MMXEXT(cpu_flags)) {
119 253 c->sad[1] = ff_sad8_mmxext;
120
121 253 c->pix_abs[1][0] = ff_sad8_mmxext;
122 253 c->pix_abs[1][1] = ff_sad8_x2_mmxext;
123 253 c->pix_abs[1][2] = ff_sad8_y2_mmxext;
124
125 253 c->vsad[5] = ff_vsad_intra8_mmxext;
126
127
2/2
✓ Branch 0 taken 232 times.
✓ Branch 1 taken 21 times.
 253 if (!(avctx->flags & AV_CODEC_FLAG_BITEXACT)) {
128 232 c->pix_abs[1][3] = ff_sad8_approx_xy2_mmxext;
129
130 232 c->vsad[1] = ff_vsad8_approx_mmxext;
131 }
132 }
133
134
2/2
✓ Branch 0 taken 251 times.
✓ Branch 1 taken 1332 times.
 1583 if (EXTERNAL_SSE2(cpu_flags)) {
135 251 c->sse[0] = ff_sse16_sse2;
136 251 c->sse[1] = ff_sse8_sse2;
137 251 c->sum_abs_dctelem = ff_sum_abs_dctelem_sse2;
138
139 251 c->pix_abs[0][0] = ff_sad16_sse2;
140 251 c->pix_abs[0][1] = ff_sad16_x2_sse2;
141 251 c->pix_abs[0][2] = ff_sad16_y2_sse2;
142 251 c->pix_abs[0][3] = ff_sad16_xy2_sse2;
143
144 251 c->hadamard8_diff[0] = ff_hadamard8_diff16_sse2;
145 251 c->hadamard8_diff[1] = ff_hadamard8_diff_sse2;
146
1/2
✓ Branch 0 taken 251 times.
✗ Branch 1 not taken.
 251 if (avctx->codec_id != AV_CODEC_ID_SNOW) {
147 251 c->sad[0] = ff_sad16_sse2;
148
149 251 c->vsad[4] = ff_vsad_intra16_sse2;
150
2/2
✓ Branch 0 taken 232 times.
✓ Branch 1 taken 19 times.
 251 if (!(avctx->flags & AV_CODEC_FLAG_BITEXACT)) {
151 232 c->vsad[0] = ff_vsad16_approx_sse2;
152 }
153 } else {
154 // Snow does not abide by the alignment requirements
155 // of blk1, so we use special versions without them for it.
156 ✗ c->sad[0] = ff_sad16u_sse2;
157
158 ✗ c->vsad[4] = ff_vsad_intra16u_sse2;
159 ✗ if (!(avctx->flags & AV_CODEC_FLAG_BITEXACT)) {
160 ✗ c->vsad[0] = ff_vsad16u_approx_sse2;
161 }
162 }
163
2/2
✓ Branch 0 taken 19 times.
✓ Branch 1 taken 232 times.
 251 if (avctx->flags & AV_CODEC_FLAG_BITEXACT) {
164 19 c->pix_abs[1][3] = ff_sad8_xy2_sse2;
165 } else {
166 232 c->pix_abs[0][3] = ff_sad16_approx_xy2_sse2;
167 }
168 }
169
170
2/2
✓ Branch 0 taken 249 times.
✓ Branch 1 taken 1334 times.
 1583 if (EXTERNAL_SSSE3(cpu_flags)) {
171 249 c->nsse[0] = nsse16_ssse3;
172 249 c->nsse[1] = nsse8_ssse3;
173
174 249 c->sum_abs_dctelem = ff_sum_abs_dctelem_ssse3;
175 249 c->hadamard8_diff[0] = ff_hadamard8_diff16_ssse3;
176 249 c->hadamard8_diff[1] = ff_hadamard8_diff_ssse3;
177
178 #if ARCH_X86_64
179 249 c->median_sad[0] = ff_median_sad16_ssse3;
180 #endif
181 249 c->median_sad[1] = ff_median_sad8_ssse3;
182 }
183 1583 }
184