1 /*
2 * Copyright (C) 2012 Michael Niedermayer (michaelni@gmx.at)
3 *
4 * This file is part of libswresample
5 *
6 * libswresample 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 * libswresample 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 libswresample; if not, write to the Free Software
18 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
19 */
20
21 #include "libavutil/attributes.h"
22 #include "libavutil/x86/cpu.h"
23 #include "libswresample/swresample_internal.h"
24 #include "libswresample/audioconvert.h"
25
26 #define PROTO(pre, in, out, cap) void ff ## pre ## in## _to_ ##out## _a_ ##cap(uint8_t **dst, const uint8_t **src, int len);
27 #define PROTO2(pre, out, cap) PROTO(pre, int16, out, cap) PROTO(pre, int32, out, cap) PROTO(pre, float, out, cap)
28 #define PROTO3(pre, cap) PROTO2(pre, int16, cap) PROTO2(pre, int32, cap) PROTO2(pre, float, cap)
29 #define PROTO4(pre) PROTO3(pre, sse) PROTO3(pre, sse2) PROTO3(pre, ssse3) PROTO3(pre, sse4) PROTO3(pre, avx) PROTO3(pre, avx2)
30 PROTO4(_)
PROTO4(_pack_2ch_)31 PROTO4(_pack_2ch_)
32 PROTO4(_pack_6ch_)
33 PROTO4(_pack_8ch_)
34 PROTO4(_unpack_2ch_)
35 PROTO4(_unpack_6ch_)
36
37 av_cold void swri_audio_convert_init_x86(struct AudioConvert *ac,
38 enum AVSampleFormat out_fmt,
39 enum AVSampleFormat in_fmt,
40 int channels){
41 int mm_flags = av_get_cpu_flags();
42
43 ac->simd_f= NULL;
44
45 //FIXME add memcpy case
46
47 #define MULTI_CAPS_FUNC(flag, cap) \
48 if (EXTERNAL_##flag(mm_flags)) {\
49 if( out_fmt == AV_SAMPLE_FMT_S32 && in_fmt == AV_SAMPLE_FMT_S16 || out_fmt == AV_SAMPLE_FMT_S32P && in_fmt == AV_SAMPLE_FMT_S16P)\
50 ac->simd_f = ff_int16_to_int32_a_ ## cap;\
51 if( out_fmt == AV_SAMPLE_FMT_S16 && in_fmt == AV_SAMPLE_FMT_S32 || out_fmt == AV_SAMPLE_FMT_S16P && in_fmt == AV_SAMPLE_FMT_S32P)\
52 ac->simd_f = ff_int32_to_int16_a_ ## cap;\
53 }
54
55 MULTI_CAPS_FUNC(SSE2, sse2)
56
57 if(EXTERNAL_SSE(mm_flags)) {
58 if(channels == 6) {
59 if( out_fmt == AV_SAMPLE_FMT_FLT && in_fmt == AV_SAMPLE_FMT_FLTP || out_fmt == AV_SAMPLE_FMT_S32 && in_fmt == AV_SAMPLE_FMT_S32P)
60 ac->simd_f = ff_pack_6ch_float_to_float_a_sse;
61
62 if( out_fmt == AV_SAMPLE_FMT_FLTP && in_fmt == AV_SAMPLE_FMT_FLT || out_fmt == AV_SAMPLE_FMT_S32P && in_fmt == AV_SAMPLE_FMT_S32)
63 ac->simd_f = ff_unpack_6ch_float_to_float_a_sse;
64 }
65 }
66 if(EXTERNAL_SSE2(mm_flags)) {
67 if( out_fmt == AV_SAMPLE_FMT_FLT && in_fmt == AV_SAMPLE_FMT_S32 || out_fmt == AV_SAMPLE_FMT_FLTP && in_fmt == AV_SAMPLE_FMT_S32P)
68 ac->simd_f = ff_int32_to_float_a_sse2;
69 if( out_fmt == AV_SAMPLE_FMT_FLT && in_fmt == AV_SAMPLE_FMT_S16 || out_fmt == AV_SAMPLE_FMT_FLTP && in_fmt == AV_SAMPLE_FMT_S16P)
70 ac->simd_f = ff_int16_to_float_a_sse2;
71 if( out_fmt == AV_SAMPLE_FMT_S32 && in_fmt == AV_SAMPLE_FMT_FLT || out_fmt == AV_SAMPLE_FMT_S32P && in_fmt == AV_SAMPLE_FMT_FLTP)
72 ac->simd_f = ff_float_to_int32_a_sse2;
73 if( out_fmt == AV_SAMPLE_FMT_S16 && in_fmt == AV_SAMPLE_FMT_FLT || out_fmt == AV_SAMPLE_FMT_S16P && in_fmt == AV_SAMPLE_FMT_FLTP)
74 ac->simd_f = ff_float_to_int16_a_sse2;
75
76 if(channels == 2) {
77 if( out_fmt == AV_SAMPLE_FMT_FLT && in_fmt == AV_SAMPLE_FMT_FLTP || out_fmt == AV_SAMPLE_FMT_S32 && in_fmt == AV_SAMPLE_FMT_S32P)
78 ac->simd_f = ff_pack_2ch_int32_to_int32_a_sse2;
79 if( out_fmt == AV_SAMPLE_FMT_S16 && in_fmt == AV_SAMPLE_FMT_S16P)
80 ac->simd_f = ff_pack_2ch_int16_to_int16_a_sse2;
81 if( out_fmt == AV_SAMPLE_FMT_S32 && in_fmt == AV_SAMPLE_FMT_S16P)
82 ac->simd_f = ff_pack_2ch_int16_to_int32_a_sse2;
83 if( out_fmt == AV_SAMPLE_FMT_S16 && in_fmt == AV_SAMPLE_FMT_S32P)
84 ac->simd_f = ff_pack_2ch_int32_to_int16_a_sse2;
85
86 if( out_fmt == AV_SAMPLE_FMT_FLTP && in_fmt == AV_SAMPLE_FMT_FLT || out_fmt == AV_SAMPLE_FMT_S32P && in_fmt == AV_SAMPLE_FMT_S32)
87 ac->simd_f = ff_unpack_2ch_int32_to_int32_a_sse2;
88 if( out_fmt == AV_SAMPLE_FMT_S16P && in_fmt == AV_SAMPLE_FMT_S16)
89 ac->simd_f = ff_unpack_2ch_int16_to_int16_a_sse2;
90 if( out_fmt == AV_SAMPLE_FMT_S32P && in_fmt == AV_SAMPLE_FMT_S16)
91 ac->simd_f = ff_unpack_2ch_int16_to_int32_a_sse2;
92 if( out_fmt == AV_SAMPLE_FMT_S16P && in_fmt == AV_SAMPLE_FMT_S32)
93 ac->simd_f = ff_unpack_2ch_int32_to_int16_a_sse2;
94
95 if( out_fmt == AV_SAMPLE_FMT_FLT && in_fmt == AV_SAMPLE_FMT_S32P)
96 ac->simd_f = ff_pack_2ch_int32_to_float_a_sse2;
97 if( out_fmt == AV_SAMPLE_FMT_S32 && in_fmt == AV_SAMPLE_FMT_FLTP)
98 ac->simd_f = ff_pack_2ch_float_to_int32_a_sse2;
99 if( out_fmt == AV_SAMPLE_FMT_FLT && in_fmt == AV_SAMPLE_FMT_S16P)
100 ac->simd_f = ff_pack_2ch_int16_to_float_a_sse2;
101 if( out_fmt == AV_SAMPLE_FMT_S16 && in_fmt == AV_SAMPLE_FMT_FLTP)
102 ac->simd_f = ff_pack_2ch_float_to_int16_a_sse2;
103 if( out_fmt == AV_SAMPLE_FMT_FLTP && in_fmt == AV_SAMPLE_FMT_S32)
104 ac->simd_f = ff_unpack_2ch_int32_to_float_a_sse2;
105 if( out_fmt == AV_SAMPLE_FMT_S32P && in_fmt == AV_SAMPLE_FMT_FLT)
106 ac->simd_f = ff_unpack_2ch_float_to_int32_a_sse2;
107 if( out_fmt == AV_SAMPLE_FMT_FLTP && in_fmt == AV_SAMPLE_FMT_S16)
108 ac->simd_f = ff_unpack_2ch_int16_to_float_a_sse2;
109 if( out_fmt == AV_SAMPLE_FMT_S16P && in_fmt == AV_SAMPLE_FMT_FLT)
110 ac->simd_f = ff_unpack_2ch_float_to_int16_a_sse2;
111 }
112 if(channels == 6) {
113 if( out_fmt == AV_SAMPLE_FMT_FLT && in_fmt == AV_SAMPLE_FMT_S32P)
114 ac->simd_f = ff_pack_6ch_int32_to_float_a_sse2;
115 if( out_fmt == AV_SAMPLE_FMT_S32 && in_fmt == AV_SAMPLE_FMT_FLTP)
116 ac->simd_f = ff_pack_6ch_float_to_int32_a_sse2;
117
118 if( out_fmt == AV_SAMPLE_FMT_FLTP && in_fmt == AV_SAMPLE_FMT_S32)
119 ac->simd_f = ff_unpack_6ch_int32_to_float_a_sse2;
120 if( out_fmt == AV_SAMPLE_FMT_S32P && in_fmt == AV_SAMPLE_FMT_FLT)
121 ac->simd_f = ff_unpack_6ch_float_to_int32_a_sse2;
122 }
123 if(channels == 8) {
124 if( out_fmt == AV_SAMPLE_FMT_FLT && in_fmt == AV_SAMPLE_FMT_FLTP || out_fmt == AV_SAMPLE_FMT_S32 && in_fmt == AV_SAMPLE_FMT_S32P)
125 ac->simd_f = ff_pack_8ch_float_to_float_a_sse2;
126 if( out_fmt == AV_SAMPLE_FMT_FLT && in_fmt == AV_SAMPLE_FMT_S32P)
127 ac->simd_f = ff_pack_8ch_int32_to_float_a_sse2;
128 if( out_fmt == AV_SAMPLE_FMT_S32 && in_fmt == AV_SAMPLE_FMT_FLTP)
129 ac->simd_f = ff_pack_8ch_float_to_int32_a_sse2;
130 }
131 }
132 if(EXTERNAL_SSSE3(mm_flags)) {
133 if(channels == 2) {
134 if( out_fmt == AV_SAMPLE_FMT_S16P && in_fmt == AV_SAMPLE_FMT_S16)
135 ac->simd_f = ff_unpack_2ch_int16_to_int16_a_ssse3;
136 if( out_fmt == AV_SAMPLE_FMT_S32P && in_fmt == AV_SAMPLE_FMT_S16)
137 ac->simd_f = ff_unpack_2ch_int16_to_int32_a_ssse3;
138 if( out_fmt == AV_SAMPLE_FMT_FLTP && in_fmt == AV_SAMPLE_FMT_S16)
139 ac->simd_f = ff_unpack_2ch_int16_to_float_a_ssse3;
140 }
141 }
142 if(EXTERNAL_AVX_FAST(mm_flags)) {
143 if( out_fmt == AV_SAMPLE_FMT_FLT && in_fmt == AV_SAMPLE_FMT_S32 || out_fmt == AV_SAMPLE_FMT_FLTP && in_fmt == AV_SAMPLE_FMT_S32P)
144 ac->simd_f = ff_int32_to_float_a_avx;
145 }
146 if(EXTERNAL_AVX(mm_flags)) {
147 if(channels == 6) {
148 if( out_fmt == AV_SAMPLE_FMT_FLT && in_fmt == AV_SAMPLE_FMT_FLTP || out_fmt == AV_SAMPLE_FMT_S32 && in_fmt == AV_SAMPLE_FMT_S32P)
149 ac->simd_f = ff_pack_6ch_float_to_float_a_avx;
150 if( out_fmt == AV_SAMPLE_FMT_FLT && in_fmt == AV_SAMPLE_FMT_S32P)
151 ac->simd_f = ff_pack_6ch_int32_to_float_a_avx;
152 if( out_fmt == AV_SAMPLE_FMT_S32 && in_fmt == AV_SAMPLE_FMT_FLTP)
153 ac->simd_f = ff_pack_6ch_float_to_int32_a_avx;
154
155 if( out_fmt == AV_SAMPLE_FMT_FLTP && in_fmt == AV_SAMPLE_FMT_FLT || out_fmt == AV_SAMPLE_FMT_S32P && in_fmt == AV_SAMPLE_FMT_S32)
156 ac->simd_f = ff_unpack_6ch_float_to_float_a_avx;
157 if( out_fmt == AV_SAMPLE_FMT_FLTP && in_fmt == AV_SAMPLE_FMT_S32)
158 ac->simd_f = ff_unpack_6ch_int32_to_float_a_avx;
159 if( out_fmt == AV_SAMPLE_FMT_S32P && in_fmt == AV_SAMPLE_FMT_FLT)
160 ac->simd_f = ff_unpack_6ch_float_to_int32_a_avx;
161 }
162 if(channels == 8) {
163 if( out_fmt == AV_SAMPLE_FMT_FLT && in_fmt == AV_SAMPLE_FMT_FLTP || out_fmt == AV_SAMPLE_FMT_S32 && in_fmt == AV_SAMPLE_FMT_S32P)
164 ac->simd_f = ff_pack_8ch_float_to_float_a_avx;
165 if( out_fmt == AV_SAMPLE_FMT_FLT && in_fmt == AV_SAMPLE_FMT_S32P)
166 ac->simd_f = ff_pack_8ch_int32_to_float_a_avx;
167 if( out_fmt == AV_SAMPLE_FMT_S32 && in_fmt == AV_SAMPLE_FMT_FLTP)
168 ac->simd_f = ff_pack_8ch_float_to_int32_a_avx;
169 }
170 }
171 if(EXTERNAL_AVX2_FAST(mm_flags)) {
172 if( out_fmt == AV_SAMPLE_FMT_S32 && in_fmt == AV_SAMPLE_FMT_FLT || out_fmt == AV_SAMPLE_FMT_S32P && in_fmt == AV_SAMPLE_FMT_FLTP)
173 ac->simd_f = ff_float_to_int32_a_avx2;
174 }
175 }
176