Line | Branch | Exec | Source |
---|---|---|---|
1 | /* | ||
2 | * x86-optimized AC-3 DSP functions | ||
3 | * Copyright (c) 2011 Justin Ruggles | ||
4 | * | ||
5 | * This file is part of FFmpeg. | ||
6 | * | ||
7 | * FFmpeg is free software; you can redistribute it and/or | ||
8 | * modify it under the terms of the GNU Lesser General Public | ||
9 | * License as published by the Free Software Foundation; either | ||
10 | * version 2.1 of the License, or (at your option) any later version. | ||
11 | * | ||
12 | * FFmpeg is distributed in the hope that it will be useful, | ||
13 | * but WITHOUT ANY WARRANTY; without even the implied warranty of | ||
14 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | ||
15 | * Lesser General Public License for more details. | ||
16 | * | ||
17 | * You should have received a copy of the GNU Lesser General Public | ||
18 | * License along with FFmpeg; if not, write to the Free Software | ||
19 | * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA | ||
20 | */ | ||
21 | |||
22 | #include "libavutil/attributes.h" | ||
23 | #include "libavutil/x86/asm.h" | ||
24 | #include "libavutil/x86/cpu.h" | ||
25 | #include "libavcodec/ac3dsp.h" | ||
26 | |||
27 | void ff_ac3_exponent_min_sse2 (uint8_t *exp, int num_reuse_blocks, int nb_coefs); | ||
28 | |||
29 | void ff_float_to_fixed24_sse2 (int32_t *dst, const float *src, size_t len); | ||
30 | void ff_float_to_fixed24_avx (int32_t *dst, const float *src, size_t len); | ||
31 | |||
32 | int ff_ac3_compute_mantissa_size_sse2(uint16_t mant_cnt[6][16]); | ||
33 | |||
34 | void ff_ac3_extract_exponents_sse2 (uint8_t *exp, int32_t *coef, int nb_coefs); | ||
35 | void ff_ac3_extract_exponents_ssse3(uint8_t *exp, int32_t *coef, int nb_coefs); | ||
36 | |||
37 | 118 | av_cold void ff_ac3dsp_init_x86(AC3DSPContext *c) | |
38 | { | ||
39 | 118 | int cpu_flags = av_get_cpu_flags(); | |
40 | |||
41 |
2/2✓ Branch 0 taken 35 times.
✓ Branch 1 taken 83 times.
|
118 | if (EXTERNAL_SSE2(cpu_flags)) { |
42 | 35 | c->ac3_exponent_min = ff_ac3_exponent_min_sse2; | |
43 | 35 | c->float_to_fixed24 = ff_float_to_fixed24_sse2; | |
44 | 35 | c->compute_mantissa_size = ff_ac3_compute_mantissa_size_sse2; | |
45 | 35 | c->extract_exponents = ff_ac3_extract_exponents_sse2; | |
46 | } | ||
47 | |||
48 |
2/2✓ Branch 0 taken 33 times.
✓ Branch 1 taken 85 times.
|
118 | if (EXTERNAL_SSSE3(cpu_flags)) { |
49 |
1/2✓ Branch 0 taken 33 times.
✗ Branch 1 not taken.
|
33 | if (!(cpu_flags & AV_CPU_FLAG_ATOM)) |
50 | 33 | c->extract_exponents = ff_ac3_extract_exponents_ssse3; | |
51 | } | ||
52 |
3/4✓ Branch 0 taken 29 times.
✓ Branch 1 taken 89 times.
✓ Branch 2 taken 29 times.
✗ Branch 3 not taken.
|
118 | if (EXTERNAL_AVX_FAST(cpu_flags)) { |
53 | 29 | c->float_to_fixed24 = ff_float_to_fixed24_avx; | |
54 | } | ||
55 | 118 | } | |
56 | |||
57 | #define DOWNMIX_FUNC_OPT(ch, opt) \ | ||
58 | void ff_ac3_downmix_ ## ch ## _to_1_ ## opt(float **samples, \ | ||
59 | float **matrix, int len); \ | ||
60 | void ff_ac3_downmix_ ## ch ## _to_2_ ## opt(float **samples, \ | ||
61 | float **matrix, int len); | ||
62 | |||
63 | #define DOWNMIX_FUNCS(opt) \ | ||
64 | DOWNMIX_FUNC_OPT(3, opt) \ | ||
65 | DOWNMIX_FUNC_OPT(4, opt) \ | ||
66 | DOWNMIX_FUNC_OPT(5, opt) \ | ||
67 | DOWNMIX_FUNC_OPT(6, opt) | ||
68 | |||
69 | DOWNMIX_FUNCS(sse) | ||
70 | DOWNMIX_FUNCS(avx) | ||
71 | DOWNMIX_FUNCS(fma3) | ||
72 | |||
73 | 11 | void ff_ac3dsp_set_downmix_x86(AC3DSPContext *c) | |
74 | { | ||
75 | 11 | int cpu_flags = av_get_cpu_flags(); | |
76 | |||
77 | #define SET_DOWNMIX(ch, suf, SUF) \ | ||
78 | if (ch == c->in_channels) { \ | ||
79 | if (EXTERNAL_ ## SUF (cpu_flags)) { \ | ||
80 | if (c->out_channels == 1) \ | ||
81 | c->downmix = ff_ac3_downmix_ ## ch ## _to_1_ ## suf; \ | ||
82 | else \ | ||
83 | c->downmix = ff_ac3_downmix_ ## ch ## _to_2_ ## suf; \ | ||
84 | } \ | ||
85 | } | ||
86 | |||
87 | #define SET_DOWNMIX_ALL(suf, SUF) \ | ||
88 | SET_DOWNMIX(3, suf, SUF) \ | ||
89 | SET_DOWNMIX(4, suf, SUF) \ | ||
90 | SET_DOWNMIX(5, suf, SUF) \ | ||
91 | SET_DOWNMIX(6, suf, SUF) | ||
92 | |||
93 |
8/24✗ Branch 0 not taken.
✓ Branch 1 taken 11 times.
✗ Branch 2 not taken.
✗ Branch 3 not taken.
✗ Branch 4 not taken.
✗ Branch 5 not taken.
✓ Branch 6 taken 5 times.
✓ Branch 7 taken 6 times.
✗ Branch 8 not taken.
✓ Branch 9 taken 5 times.
✗ Branch 10 not taken.
✗ Branch 11 not taken.
✓ Branch 12 taken 6 times.
✓ Branch 13 taken 5 times.
✗ Branch 14 not taken.
✓ Branch 15 taken 6 times.
✗ Branch 16 not taken.
✗ Branch 17 not taken.
✗ Branch 18 not taken.
✓ Branch 19 taken 11 times.
✗ Branch 20 not taken.
✗ Branch 21 not taken.
✗ Branch 22 not taken.
✗ Branch 23 not taken.
|
11 | SET_DOWNMIX_ALL(sse, SSE) |
94 |
1/2✓ Branch 0 taken 11 times.
✗ Branch 1 not taken.
|
11 | if (!(cpu_flags & AV_CPU_FLAG_AVXSLOW)) { |
95 |
8/24✗ Branch 0 not taken.
✓ Branch 1 taken 11 times.
✗ Branch 2 not taken.
✗ Branch 3 not taken.
✗ Branch 4 not taken.
✗ Branch 5 not taken.
✓ Branch 6 taken 5 times.
✓ Branch 7 taken 6 times.
✗ Branch 8 not taken.
✓ Branch 9 taken 5 times.
✗ Branch 10 not taken.
✗ Branch 11 not taken.
✓ Branch 12 taken 6 times.
✓ Branch 13 taken 5 times.
✗ Branch 14 not taken.
✓ Branch 15 taken 6 times.
✗ Branch 16 not taken.
✗ Branch 17 not taken.
✗ Branch 18 not taken.
✓ Branch 19 taken 11 times.
✗ Branch 20 not taken.
✗ Branch 21 not taken.
✗ Branch 22 not taken.
✗ Branch 23 not taken.
|
11 | SET_DOWNMIX_ALL(avx, AVX) |
96 |
8/24✗ Branch 0 not taken.
✓ Branch 1 taken 11 times.
✗ Branch 2 not taken.
✗ Branch 3 not taken.
✗ Branch 4 not taken.
✗ Branch 5 not taken.
✓ Branch 6 taken 5 times.
✓ Branch 7 taken 6 times.
✗ Branch 8 not taken.
✓ Branch 9 taken 5 times.
✗ Branch 10 not taken.
✗ Branch 11 not taken.
✓ Branch 12 taken 6 times.
✓ Branch 13 taken 5 times.
✗ Branch 14 not taken.
✓ Branch 15 taken 6 times.
✗ Branch 16 not taken.
✗ Branch 17 not taken.
✗ Branch 18 not taken.
✓ Branch 19 taken 11 times.
✗ Branch 20 not taken.
✗ Branch 21 not taken.
✗ Branch 22 not taken.
✗ Branch 23 not taken.
|
11 | SET_DOWNMIX_ALL(fma3, FMA3) |
97 | } | ||
98 | 11 | } | |
99 |