x86: check for AV_CPU_FLAG_AVXSLOW where useful
[libav.git] / libavresample / x86 / audio_convert_init.c
1 /*
2 * Copyright (c) 2012 Justin Ruggles <justin.ruggles@gmail.com>
3 *
4 * This file is part of Libav.
5 *
6 * Libav is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU Lesser General Public
8 * License as published by the Free Software Foundation; either
9 * version 2.1 of the License, or (at your option) any later version.
10 *
11 * Libav 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 GNU
14 * Lesser General Public License for more details.
15 *
16 * You should have received a copy of the GNU Lesser General Public
17 * License along with Libav; if not, write to the Free Software
18 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
19 */
20
21 #include "config.h"
22 #include "libavutil/cpu.h"
23 #include "libavutil/x86/cpu.h"
24 #include "libavresample/audio_convert.h"
25
26 /* flat conversions */
27
28 void ff_conv_s16_to_s32_sse2(int16_t *dst, const int32_t *src, int len);
29
30 void ff_conv_s16_to_flt_sse2(float *dst, const int16_t *src, int len);
31 void ff_conv_s16_to_flt_sse4(float *dst, const int16_t *src, int len);
32
33 void ff_conv_s32_to_s16_mmx (int16_t *dst, const int32_t *src, int len);
34 void ff_conv_s32_to_s16_sse2(int16_t *dst, const int32_t *src, int len);
35
36 void ff_conv_s32_to_flt_sse2(float *dst, const int32_t *src, int len);
37 void ff_conv_s32_to_flt_avx (float *dst, const int32_t *src, int len);
38
39 void ff_conv_flt_to_s16_sse2(int16_t *dst, const float *src, int len);
40
41 void ff_conv_flt_to_s32_sse2(int32_t *dst, const float *src, int len);
42 void ff_conv_flt_to_s32_avx (int32_t *dst, const float *src, int len);
43
44 /* interleave conversions */
45
46 void ff_conv_s16p_to_s16_2ch_sse2(int16_t *dst, int16_t *const *src,
47 int len, int channels);
48 void ff_conv_s16p_to_s16_2ch_avx (int16_t *dst, int16_t *const *src,
49 int len, int channels);
50
51 void ff_conv_s16p_to_s16_6ch_sse2(int16_t *dst, int16_t *const *src,
52 int len, int channels);
53 void ff_conv_s16p_to_s16_6ch_sse2slow(int16_t *dst, int16_t *const *src,
54 int len, int channels);
55 void ff_conv_s16p_to_s16_6ch_avx (int16_t *dst, int16_t *const *src,
56 int len, int channels);
57
58 void ff_conv_s16p_to_flt_2ch_sse2(float *dst, int16_t *const *src,
59 int len, int channels);
60 void ff_conv_s16p_to_flt_2ch_avx (float *dst, int16_t *const *src,
61 int len, int channels);
62
63 void ff_conv_s16p_to_flt_6ch_sse2 (float *dst, int16_t *const *src,
64 int len, int channels);
65 void ff_conv_s16p_to_flt_6ch_ssse3(float *dst, int16_t *const *src,
66 int len, int channels);
67 void ff_conv_s16p_to_flt_6ch_avx (float *dst, int16_t *const *src,
68 int len, int channels);
69
70 void ff_conv_fltp_to_s16_2ch_sse2 (int16_t *dst, float *const *src,
71 int len, int channels);
72 void ff_conv_fltp_to_s16_2ch_ssse3(int16_t *dst, float *const *src,
73 int len, int channels);
74
75 void ff_conv_fltp_to_s16_6ch_sse (int16_t *dst, float *const *src,
76 int len, int channels);
77 void ff_conv_fltp_to_s16_6ch_sse2(int16_t *dst, float *const *src,
78 int len, int channels);
79 void ff_conv_fltp_to_s16_6ch_avx (int16_t *dst, float *const *src,
80 int len, int channels);
81
82 void ff_conv_fltp_to_flt_2ch_sse(float *dst, float *const *src, int len,
83 int channels);
84 void ff_conv_fltp_to_flt_2ch_avx(float *dst, float *const *src, int len,
85 int channels);
86
87 void ff_conv_fltp_to_flt_6ch_mmx (float *dst, float *const *src, int len,
88 int channels);
89 void ff_conv_fltp_to_flt_6ch_sse4(float *dst, float *const *src, int len,
90 int channels);
91 void ff_conv_fltp_to_flt_6ch_avx (float *dst, float *const *src, int len,
92 int channels);
93
94 /* deinterleave conversions */
95
96 void ff_conv_s16_to_s16p_2ch_sse2(int16_t *const *dst, int16_t *src,
97 int len, int channels);
98 void ff_conv_s16_to_s16p_2ch_ssse3(int16_t *const *dst, int16_t *src,
99 int len, int channels);
100 void ff_conv_s16_to_s16p_2ch_avx (int16_t *const *dst, int16_t *src,
101 int len, int channels);
102
103 void ff_conv_s16_to_s16p_6ch_sse2 (int16_t *const *dst, int16_t *src,
104 int len, int channels);
105 void ff_conv_s16_to_s16p_6ch_ssse3(int16_t *const *dst, int16_t *src,
106 int len, int channels);
107 void ff_conv_s16_to_s16p_6ch_avx (int16_t *const *dst, int16_t *src,
108 int len, int channels);
109
110 void ff_conv_s16_to_fltp_2ch_sse2(float *const *dst, int16_t *src,
111 int len, int channels);
112 void ff_conv_s16_to_fltp_2ch_avx (float *const *dst, int16_t *src,
113 int len, int channels);
114
115 void ff_conv_s16_to_fltp_6ch_sse2 (float *const *dst, int16_t *src,
116 int len, int channels);
117 void ff_conv_s16_to_fltp_6ch_ssse3(float *const *dst, int16_t *src,
118 int len, int channels);
119 void ff_conv_s16_to_fltp_6ch_sse4 (float *const *dst, int16_t *src,
120 int len, int channels);
121 void ff_conv_s16_to_fltp_6ch_avx (float *const *dst, int16_t *src,
122 int len, int channels);
123
124 void ff_conv_flt_to_s16p_2ch_sse2(int16_t *const *dst, float *src,
125 int len, int channels);
126 void ff_conv_flt_to_s16p_2ch_avx (int16_t *const *dst, float *src,
127 int len, int channels);
128
129 void ff_conv_flt_to_s16p_6ch_sse2 (int16_t *const *dst, float *src,
130 int len, int channels);
131 void ff_conv_flt_to_s16p_6ch_ssse3(int16_t *const *dst, float *src,
132 int len, int channels);
133 void ff_conv_flt_to_s16p_6ch_avx (int16_t *const *dst, float *src,
134 int len, int channels);
135
136 void ff_conv_flt_to_fltp_2ch_sse(float *const *dst, float *src, int len,
137 int channels);
138 void ff_conv_flt_to_fltp_2ch_avx(float *const *dst, float *src, int len,
139 int channels);
140
141 void ff_conv_flt_to_fltp_6ch_sse2(float *const *dst, float *src, int len,
142 int channels);
143 void ff_conv_flt_to_fltp_6ch_avx (float *const *dst, float *src, int len,
144 int channels);
145
146 av_cold void ff_audio_convert_init_x86(AudioConvert *ac)
147 {
148 int cpu_flags = av_get_cpu_flags();
149
150 if (EXTERNAL_MMX(cpu_flags)) {
151 ff_audio_convert_set_func(ac, AV_SAMPLE_FMT_S16, AV_SAMPLE_FMT_S32,
152 0, 1, 8, "MMX", ff_conv_s32_to_s16_mmx);
153 ff_audio_convert_set_func(ac, AV_SAMPLE_FMT_FLT, AV_SAMPLE_FMT_FLTP,
154 6, 1, 4, "MMX", ff_conv_fltp_to_flt_6ch_mmx);
155 }
156 if (EXTERNAL_SSE(cpu_flags)) {
157 ff_audio_convert_set_func(ac, AV_SAMPLE_FMT_S16, AV_SAMPLE_FMT_FLTP,
158 6, 1, 2, "SSE", ff_conv_fltp_to_s16_6ch_sse);
159 ff_audio_convert_set_func(ac, AV_SAMPLE_FMT_FLT, AV_SAMPLE_FMT_FLTP,
160 2, 16, 8, "SSE", ff_conv_fltp_to_flt_2ch_sse);
161 ff_audio_convert_set_func(ac, AV_SAMPLE_FMT_FLTP, AV_SAMPLE_FMT_FLT,
162 2, 16, 4, "SSE", ff_conv_flt_to_fltp_2ch_sse);
163 }
164 if (EXTERNAL_SSE2(cpu_flags)) {
165 if (!(cpu_flags & AV_CPU_FLAG_SSE2SLOW)) {
166 ff_audio_convert_set_func(ac, AV_SAMPLE_FMT_S16, AV_SAMPLE_FMT_S32,
167 0, 16, 16, "SSE2", ff_conv_s32_to_s16_sse2);
168 ff_audio_convert_set_func(ac, AV_SAMPLE_FMT_S16, AV_SAMPLE_FMT_S16P,
169 6, 16, 8, "SSE2", ff_conv_s16p_to_s16_6ch_sse2);
170 ff_audio_convert_set_func(ac, AV_SAMPLE_FMT_S16, AV_SAMPLE_FMT_FLTP,
171 6, 16, 4, "SSE2", ff_conv_fltp_to_s16_6ch_sse2);
172 } else {
173 ff_audio_convert_set_func(ac, AV_SAMPLE_FMT_S16, AV_SAMPLE_FMT_S16P,
174 6, 1, 4, "SSE2SLOW", ff_conv_s16p_to_s16_6ch_sse2slow);
175 }
176 ff_audio_convert_set_func(ac, AV_SAMPLE_FMT_S32, AV_SAMPLE_FMT_S16,
177 0, 16, 8, "SSE2", ff_conv_s16_to_s32_sse2);
178 ff_audio_convert_set_func(ac, AV_SAMPLE_FMT_FLT, AV_SAMPLE_FMT_S16,
179 0, 16, 8, "SSE2", ff_conv_s16_to_flt_sse2);
180 ff_audio_convert_set_func(ac, AV_SAMPLE_FMT_FLT, AV_SAMPLE_FMT_S32,
181 0, 16, 8, "SSE2", ff_conv_s32_to_flt_sse2);
182 ff_audio_convert_set_func(ac, AV_SAMPLE_FMT_S16, AV_SAMPLE_FMT_FLT,
183 0, 16, 16, "SSE2", ff_conv_flt_to_s16_sse2);
184 ff_audio_convert_set_func(ac, AV_SAMPLE_FMT_S32, AV_SAMPLE_FMT_FLT,
185 0, 16, 16, "SSE2", ff_conv_flt_to_s32_sse2);
186 ff_audio_convert_set_func(ac, AV_SAMPLE_FMT_S16, AV_SAMPLE_FMT_S16P,
187 2, 16, 16, "SSE2", ff_conv_s16p_to_s16_2ch_sse2);
188 ff_audio_convert_set_func(ac, AV_SAMPLE_FMT_FLT, AV_SAMPLE_FMT_S16P,
189 2, 16, 8, "SSE2", ff_conv_s16p_to_flt_2ch_sse2);
190 ff_audio_convert_set_func(ac, AV_SAMPLE_FMT_FLT, AV_SAMPLE_FMT_S16P,
191 6, 16, 4, "SSE2", ff_conv_s16p_to_flt_6ch_sse2);
192 ff_audio_convert_set_func(ac, AV_SAMPLE_FMT_S16, AV_SAMPLE_FMT_FLTP,
193 2, 16, 4, "SSE2", ff_conv_fltp_to_s16_2ch_sse2);
194 ff_audio_convert_set_func(ac, AV_SAMPLE_FMT_S16P, AV_SAMPLE_FMT_S16,
195 2, 16, 8, "SSE2", ff_conv_s16_to_s16p_2ch_sse2);
196 ff_audio_convert_set_func(ac, AV_SAMPLE_FMT_S16P, AV_SAMPLE_FMT_S16,
197 6, 16, 4, "SSE2", ff_conv_s16_to_s16p_6ch_sse2);
198 ff_audio_convert_set_func(ac, AV_SAMPLE_FMT_FLTP, AV_SAMPLE_FMT_S16,
199 2, 16, 8, "SSE2", ff_conv_s16_to_fltp_2ch_sse2);
200 ff_audio_convert_set_func(ac, AV_SAMPLE_FMT_FLTP, AV_SAMPLE_FMT_S16,
201 6, 16, 4, "SSE2", ff_conv_s16_to_fltp_6ch_sse2);
202 ff_audio_convert_set_func(ac, AV_SAMPLE_FMT_S16P, AV_SAMPLE_FMT_FLT,
203 2, 16, 8, "SSE2", ff_conv_flt_to_s16p_2ch_sse2);
204 ff_audio_convert_set_func(ac, AV_SAMPLE_FMT_S16P, AV_SAMPLE_FMT_FLT,
205 6, 16, 4, "SSE2", ff_conv_flt_to_s16p_6ch_sse2);
206 ff_audio_convert_set_func(ac, AV_SAMPLE_FMT_FLTP, AV_SAMPLE_FMT_FLT,
207 6, 16, 4, "SSE2", ff_conv_flt_to_fltp_6ch_sse2);
208 }
209 if (EXTERNAL_SSSE3(cpu_flags)) {
210 ff_audio_convert_set_func(ac, AV_SAMPLE_FMT_FLT, AV_SAMPLE_FMT_S16P,
211 6, 16, 4, "SSSE3", ff_conv_s16p_to_flt_6ch_ssse3);
212 ff_audio_convert_set_func(ac, AV_SAMPLE_FMT_S16, AV_SAMPLE_FMT_FLTP,
213 2, 16, 4, "SSSE3", ff_conv_fltp_to_s16_2ch_ssse3);
214 ff_audio_convert_set_func(ac, AV_SAMPLE_FMT_S16P, AV_SAMPLE_FMT_S16,
215 2, 16, 8, "SSSE3", ff_conv_s16_to_s16p_2ch_ssse3);
216 ff_audio_convert_set_func(ac, AV_SAMPLE_FMT_S16P, AV_SAMPLE_FMT_S16,
217 6, 16, 4, "SSSE3", ff_conv_s16_to_s16p_6ch_ssse3);
218 ff_audio_convert_set_func(ac, AV_SAMPLE_FMT_FLTP, AV_SAMPLE_FMT_S16,
219 6, 16, 4, "SSSE3", ff_conv_s16_to_fltp_6ch_ssse3);
220 ff_audio_convert_set_func(ac, AV_SAMPLE_FMT_S16P, AV_SAMPLE_FMT_FLT,
221 6, 16, 4, "SSSE3", ff_conv_flt_to_s16p_6ch_ssse3);
222 }
223 if (EXTERNAL_SSE4(cpu_flags)) {
224 ff_audio_convert_set_func(ac, AV_SAMPLE_FMT_FLT, AV_SAMPLE_FMT_S16,
225 0, 16, 8, "SSE4", ff_conv_s16_to_flt_sse4);
226 ff_audio_convert_set_func(ac, AV_SAMPLE_FMT_FLT, AV_SAMPLE_FMT_FLTP,
227 6, 16, 4, "SSE4", ff_conv_fltp_to_flt_6ch_sse4);
228 }
229 if (EXTERNAL_AVX_FAST(cpu_flags)) {
230 ff_audio_convert_set_func(ac, AV_SAMPLE_FMT_FLT, AV_SAMPLE_FMT_S32,
231 0, 32, 16, "AVX", ff_conv_s32_to_flt_avx);
232 ff_audio_convert_set_func(ac, AV_SAMPLE_FMT_S32, AV_SAMPLE_FMT_FLT,
233 0, 32, 32, "AVX", ff_conv_flt_to_s32_avx);
234 }
235 if (EXTERNAL_AVX(cpu_flags)) {
236 ff_audio_convert_set_func(ac, AV_SAMPLE_FMT_S16, AV_SAMPLE_FMT_S16P,
237 2, 16, 16, "AVX", ff_conv_s16p_to_s16_2ch_avx);
238 ff_audio_convert_set_func(ac, AV_SAMPLE_FMT_S16, AV_SAMPLE_FMT_S16P,
239 6, 16, 8, "AVX", ff_conv_s16p_to_s16_6ch_avx);
240 ff_audio_convert_set_func(ac, AV_SAMPLE_FMT_FLT, AV_SAMPLE_FMT_S16P,
241 2, 16, 8, "AVX", ff_conv_s16p_to_flt_2ch_avx);
242 ff_audio_convert_set_func(ac, AV_SAMPLE_FMT_FLT, AV_SAMPLE_FMT_S16P,
243 6, 16, 4, "AVX", ff_conv_s16p_to_flt_6ch_avx);
244 ff_audio_convert_set_func(ac, AV_SAMPLE_FMT_S16, AV_SAMPLE_FMT_FLTP,
245 6, 16, 4, "AVX", ff_conv_fltp_to_s16_6ch_avx);
246 ff_audio_convert_set_func(ac, AV_SAMPLE_FMT_FLT, AV_SAMPLE_FMT_FLTP,
247 6, 16, 4, "AVX", ff_conv_fltp_to_flt_6ch_avx);
248 ff_audio_convert_set_func(ac, AV_SAMPLE_FMT_S16P, AV_SAMPLE_FMT_S16,
249 2, 16, 8, "AVX", ff_conv_s16_to_s16p_2ch_avx);
250 ff_audio_convert_set_func(ac, AV_SAMPLE_FMT_S16P, AV_SAMPLE_FMT_S16,
251 6, 16, 4, "AVX", ff_conv_s16_to_s16p_6ch_avx);
252 ff_audio_convert_set_func(ac, AV_SAMPLE_FMT_FLTP, AV_SAMPLE_FMT_S16,
253 2, 16, 8, "AVX", ff_conv_s16_to_fltp_2ch_avx);
254 ff_audio_convert_set_func(ac, AV_SAMPLE_FMT_FLTP, AV_SAMPLE_FMT_S16,
255 6, 16, 4, "AVX", ff_conv_s16_to_fltp_6ch_avx);
256 ff_audio_convert_set_func(ac, AV_SAMPLE_FMT_S16P, AV_SAMPLE_FMT_FLT,
257 2, 16, 8, "AVX", ff_conv_flt_to_s16p_2ch_avx);
258 ff_audio_convert_set_func(ac, AV_SAMPLE_FMT_S16P, AV_SAMPLE_FMT_FLT,
259 6, 16, 4, "AVX", ff_conv_flt_to_s16p_6ch_avx);
260 ff_audio_convert_set_func(ac, AV_SAMPLE_FMT_FLTP, AV_SAMPLE_FMT_FLT,
261 2, 16, 4, "AVX", ff_conv_flt_to_fltp_2ch_avx);
262 ff_audio_convert_set_func(ac, AV_SAMPLE_FMT_FLTP, AV_SAMPLE_FMT_FLT,
263 6, 16, 4, "AVX", ff_conv_flt_to_fltp_6ch_avx);
264 }
265 }