2
* Format Conversion Utils
3
* Copyright (c) 2000, 2001 Fabrice Bellard
4
* Copyright (c) 2002-2004 Michael Niedermayer <michaelni@gmx.at>
6
* This file is part of Libav.
8
* Libav is free software; you can redistribute it and/or
9
* modify it under the terms of the GNU Lesser General Public
10
* License as published by the Free Software Foundation; either
11
* version 2.1 of the License, or (at your option) any later version.
13
* Libav is distributed in the hope that it will be useful,
14
* but WITHOUT ANY WARRANTY; without even the implied warranty of
15
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
16
* Lesser General Public License for more details.
18
* You should have received a copy of the GNU Lesser General Public
19
* License along with Libav; if not, write to the Free Software
20
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
22
* MMX optimization by Nick Kurshev <nickols_k@mail.ru>
25
#include "libavutil/cpu.h"
26
#include "libavutil/x86_cpu.h"
27
#include "libavcodec/fmtconvert.h"
31
void ff_int32_to_float_fmul_scalar_sse (float *dst, const int *src, float mul, int len);
32
void ff_int32_to_float_fmul_scalar_sse2(float *dst, const int *src, float mul, int len);
34
void ff_float_to_int16_3dnow(int16_t *dst, const float *src, long len);
35
void ff_float_to_int16_sse (int16_t *dst, const float *src, long len);
36
void ff_float_to_int16_sse2 (int16_t *dst, const float *src, long len);
38
void ff_float_to_int16_interleave2_3dnow(int16_t *dst, const float **src, long len);
39
void ff_float_to_int16_interleave2_sse (int16_t *dst, const float **src, long len);
40
void ff_float_to_int16_interleave2_sse2 (int16_t *dst, const float **src, long len);
42
void ff_float_to_int16_interleave6_sse(int16_t *dst, const float **src, int len);
43
void ff_float_to_int16_interleave6_3dnow(int16_t *dst, const float **src, int len);
44
void ff_float_to_int16_interleave6_3dn2(int16_t *dst, const float **src, int len);
46
#define ff_float_to_int16_interleave6_sse2 ff_float_to_int16_interleave6_sse
48
#define FLOAT_TO_INT16_INTERLEAVE(cpu) \
49
/* gcc pessimizes register allocation if this is in the same function as float_to_int16_interleave_sse2*/\
50
static av_noinline void float_to_int16_interleave_misc_##cpu(int16_t *dst, const float **src, long len, int channels){\
51
DECLARE_ALIGNED(16, int16_t, tmp)[len];\
53
for(c=0; c<channels; c++){\
54
ff_float_to_int16_##cpu(tmp, src[c], len);\
55
for(i=0, j=c; i<len; i++, j+=channels)\
60
static void float_to_int16_interleave_##cpu(int16_t *dst, const float **src, long len, int channels){\
62
ff_float_to_int16_##cpu(dst, src[0], len);\
63
else if(channels==2){\
64
ff_float_to_int16_interleave2_##cpu(dst, src, len);\
65
}else if(channels==6){\
66
ff_float_to_int16_interleave6_##cpu(dst, src, len);\
68
float_to_int16_interleave_misc_##cpu(dst, src, len, channels);\
71
FLOAT_TO_INT16_INTERLEAVE(3dnow)
72
FLOAT_TO_INT16_INTERLEAVE(sse)
73
FLOAT_TO_INT16_INTERLEAVE(sse2)
75
static void float_to_int16_interleave_3dn2(int16_t *dst, const float **src, long len, int channels){
77
ff_float_to_int16_interleave6_3dn2(dst, src, len);
79
float_to_int16_interleave_3dnow(dst, src, len, channels);
82
void ff_float_interleave2_mmx(float *dst, const float **src, unsigned int len);
83
void ff_float_interleave2_sse(float *dst, const float **src, unsigned int len);
85
void ff_float_interleave6_mmx(float *dst, const float **src, unsigned int len);
86
void ff_float_interleave6_sse(float *dst, const float **src, unsigned int len);
88
static void float_interleave_mmx(float *dst, const float **src,
89
unsigned int len, int channels)
92
ff_float_interleave2_mmx(dst, src, len);
93
} else if (channels == 6)
94
ff_float_interleave6_mmx(dst, src, len);
96
ff_float_interleave_c(dst, src, len, channels);
99
static void float_interleave_sse(float *dst, const float **src,
100
unsigned int len, int channels)
103
ff_float_interleave2_sse(dst, src, len);
104
} else if (channels == 6)
105
ff_float_interleave6_sse(dst, src, len);
107
ff_float_interleave_c(dst, src, len, channels);
111
void ff_fmt_convert_init_x86(FmtConvertContext *c, AVCodecContext *avctx)
113
int mm_flags = av_get_cpu_flags();
116
if (mm_flags & AV_CPU_FLAG_MMX) {
117
c->float_interleave = float_interleave_mmx;
119
if (HAVE_AMD3DNOW && mm_flags & AV_CPU_FLAG_3DNOW) {
120
if(!(avctx->flags & CODEC_FLAG_BITEXACT)){
121
c->float_to_int16 = ff_float_to_int16_3dnow;
122
c->float_to_int16_interleave = float_to_int16_interleave_3dnow;
125
if (HAVE_AMD3DNOWEXT && mm_flags & AV_CPU_FLAG_3DNOWEXT) {
126
if(!(avctx->flags & CODEC_FLAG_BITEXACT)){
127
c->float_to_int16_interleave = float_to_int16_interleave_3dn2;
130
if (HAVE_SSE && mm_flags & AV_CPU_FLAG_SSE) {
131
c->int32_to_float_fmul_scalar = ff_int32_to_float_fmul_scalar_sse;
132
c->float_to_int16 = ff_float_to_int16_sse;
133
c->float_to_int16_interleave = float_to_int16_interleave_sse;
134
c->float_interleave = float_interleave_sse;
136
if (HAVE_SSE && mm_flags & AV_CPU_FLAG_SSE2) {
137
c->int32_to_float_fmul_scalar = ff_int32_to_float_fmul_scalar_sse2;
138
c->float_to_int16 = ff_float_to_int16_sse2;
139
c->float_to_int16_interleave = float_to_int16_interleave_sse2;