1 /*
2 * Format Conversion Utils
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/asm.h"
28 #include "libavutil/x86/cpu.h"
29 #include "libavcodec/fmtconvert.h"
30
31 #if HAVE_YASM
32
33 void ff_int32_to_float_fmul_scalar_sse (float *dst, const int32_t *src, float mul, int len);
34 void ff_int32_to_float_fmul_scalar_sse2(float *dst, const int32_t *src, float mul, int len);
35
36 void ff_float_to_int16_3dnow(int16_t *dst, const float *src, long len);
37 void ff_float_to_int16_sse (int16_t *dst, const float *src, long len);
38 void ff_float_to_int16_sse2 (int16_t *dst, const float *src, long len);
39
40 void ff_float_to_int16_step_3dnow(int16_t *dst, const float *src, long len, long step);
41 void ff_float_to_int16_step_sse (int16_t *dst, const float *src, long len, long step);
42 void ff_float_to_int16_step_sse2 (int16_t *dst, const float *src, long len, long step);
43
44 void ff_float_to_int16_interleave2_3dnow(int16_t *dst, const float **src, long len);
45 void ff_float_to_int16_interleave2_sse (int16_t *dst, const float **src, long len);
46 void ff_float_to_int16_interleave2_sse2 (int16_t *dst, const float **src, long len);
47
48 void ff_float_to_int16_interleave6_sse(int16_t *dst, const float **src, int len);
49 void ff_float_to_int16_interleave6_3dnow(int16_t *dst, const float **src, int len);
50 void ff_float_to_int16_interleave6_3dnowext(int16_t *dst, const float **src, int len);
51
52 #define ff_float_to_int16_interleave6_sse2 ff_float_to_int16_interleave6_sse
53
54 #define FLOAT_TO_INT16_INTERLEAVE(cpu) \
55 /* gcc pessimizes register allocation if this is in the same function as float_to_int16_interleave_sse2*/\
56 static av_noinline void float_to_int16_interleave_misc_##cpu(int16_t *dst, const float **src, long len, int channels){\
57 int c;\
58 for(c=0; c<channels; c++){\
59 ff_float_to_int16_step_##cpu(dst+c, src[c], len, channels);\
60 }\
61 }\
62 \
63 static void float_to_int16_interleave_##cpu(int16_t *dst, const float **src, long len, int channels){\
64 if(channels==1)\
65 ff_float_to_int16_##cpu(dst, src[0], len);\
66 else if(channels==2){\
67 ff_float_to_int16_interleave2_##cpu(dst, src, len);\
68 }else if(channels==6){\
69 ff_float_to_int16_interleave6_##cpu(dst, src, len);\
70 }else\
71 float_to_int16_interleave_misc_##cpu(dst, src, len, channels);\
72 }
73
74 FLOAT_TO_INT16_INTERLEAVE(3dnow)
FLOAT_TO_INT16_INTERLEAVE(sse)75 FLOAT_TO_INT16_INTERLEAVE(sse)
76 FLOAT_TO_INT16_INTERLEAVE(sse2)
77
78 static void float_to_int16_interleave_3dnowext(int16_t *dst, const float **src,
79 long len, int channels)
80 {
81 if(channels==6)
82 ff_float_to_int16_interleave6_3dnowext(dst, src, len);
83 else
84 float_to_int16_interleave_3dnow(dst, src, len, channels);
85 }
86
87 void ff_float_interleave2_mmx(float *dst, const float **src, unsigned int len);
88 void ff_float_interleave2_sse(float *dst, const float **src, unsigned int len);
89
90 void ff_float_interleave6_mmx(float *dst, const float **src, unsigned int len);
91 void ff_float_interleave6_sse(float *dst, const float **src, unsigned int len);
92
float_interleave_mmx(float * dst,const float ** src,unsigned int len,int channels)93 static void float_interleave_mmx(float *dst, const float **src,
94 unsigned int len, int channels)
95 {
96 if (channels == 2) {
97 ff_float_interleave2_mmx(dst, src, len);
98 } else if (channels == 6)
99 ff_float_interleave6_mmx(dst, src, len);
100 else
101 ff_float_interleave_c(dst, src, len, channels);
102 }
103
float_interleave_sse(float * dst,const float ** src,unsigned int len,int channels)104 static void float_interleave_sse(float *dst, const float **src,
105 unsigned int len, int channels)
106 {
107 if (channels == 2) {
108 ff_float_interleave2_sse(dst, src, len);
109 } else if (channels == 6)
110 ff_float_interleave6_sse(dst, src, len);
111 else
112 ff_float_interleave_c(dst, src, len, channels);
113 }
114 #endif /* HAVE_YASM */
115
ff_fmt_convert_init_x86(FmtConvertContext * c,AVCodecContext * avctx)116 av_cold void ff_fmt_convert_init_x86(FmtConvertContext *c, AVCodecContext *avctx)
117 {
118 #if HAVE_YASM
119 int cpu_flags = av_get_cpu_flags();
120
121 if (EXTERNAL_MMX(cpu_flags)) {
122 c->float_interleave = float_interleave_mmx;
123 }
124 if (EXTERNAL_AMD3DNOW(cpu_flags)) {
125 if (!(avctx->flags & CODEC_FLAG_BITEXACT)) {
126 c->float_to_int16 = ff_float_to_int16_3dnow;
127 c->float_to_int16_interleave = float_to_int16_interleave_3dnow;
128 }
129 }
130 if (EXTERNAL_AMD3DNOWEXT(cpu_flags)) {
131 if (!(avctx->flags & CODEC_FLAG_BITEXACT)) {
132 c->float_to_int16_interleave = float_to_int16_interleave_3dnowext;
133 }
134 }
135 if (EXTERNAL_SSE(cpu_flags)) {
136 c->int32_to_float_fmul_scalar = ff_int32_to_float_fmul_scalar_sse;
137 c->float_to_int16 = ff_float_to_int16_sse;
138 c->float_to_int16_interleave = float_to_int16_interleave_sse;
139 c->float_interleave = float_interleave_sse;
140 }
141 if (EXTERNAL_SSE2(cpu_flags)) {
142 c->int32_to_float_fmul_scalar = ff_int32_to_float_fmul_scalar_sse2;
143 c->float_to_int16 = ff_float_to_int16_sse2;
144 c->float_to_int16_interleave = float_to_int16_interleave_sse2;
145 }
146 #endif /* HAVE_YASM */
147 }
148