fmtconvert_init.c 5.81 KB
Newer Older
1 2 3 4 5
/*
 * Format Conversion Utils
 * Copyright (c) 2000, 2001 Fabrice Bellard
 * Copyright (c) 2002-2004 Michael Niedermayer <michaelni@gmx.at>
 *
6 7
 * MMX optimization by Nick Kurshev <nickols_k@mail.ru>
 *
8
 * This file is part of Libav.
9
 *
10
 * Libav is free software; you can redistribute it and/or
11 12 13 14
 * modify it under the terms of the GNU Lesser General Public
 * License as published by the Free Software Foundation; either
 * version 2.1 of the License, or (at your option) any later version.
 *
15
 * Libav is distributed in the hope that it will be useful,
16 17 18 19 20
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * Lesser General Public License for more details.
 *
 * You should have received a copy of the GNU Lesser General Public
21
 * License along with Libav; if not, write to the Free Software
22 23 24
 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
 */

25
#include "libavutil/attributes.h"
26
#include "libavutil/cpu.h"
27
#include "libavutil/x86/asm.h"
28
#include "libavutil/x86/cpu.h"
29 30
#include "libavcodec/fmtconvert.h"

31
#if HAVE_YASM
32

33 34
void ff_int32_to_float_fmul_scalar_sse (float *dst, const int32_t *src, float mul, int len);
void ff_int32_to_float_fmul_scalar_sse2(float *dst, const int32_t *src, float mul, int len);
35

36 37 38
void ff_float_to_int16_3dnow(int16_t *dst, const float *src, long len);
void ff_float_to_int16_sse  (int16_t *dst, const float *src, long len);
void ff_float_to_int16_sse2 (int16_t *dst, const float *src, long len);
39

40 41 42 43
void ff_float_to_int16_step_3dnow(int16_t *dst, const float *src, long len, long step);
void ff_float_to_int16_step_sse  (int16_t *dst, const float *src, long len, long step);
void ff_float_to_int16_step_sse2 (int16_t *dst, const float *src, long len, long step);

44 45 46 47
void ff_float_to_int16_interleave2_3dnow(int16_t *dst, const float **src, long len);
void ff_float_to_int16_interleave2_sse  (int16_t *dst, const float **src, long len);
void ff_float_to_int16_interleave2_sse2 (int16_t *dst, const float **src, long len);

48 49
void ff_float_to_int16_interleave6_sse(int16_t *dst, const float **src, int len);
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 53

#define ff_float_to_int16_interleave6_sse2 ff_float_to_int16_interleave6_sse

54
#define FLOAT_TO_INT16_INTERLEAVE(cpu) \
55 56
/* gcc pessimizes register allocation if this is in the same function as float_to_int16_interleave_sse2*/\
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 64
    }\
}\
\
static void float_to_int16_interleave_##cpu(int16_t *dst, const float **src, long len, int channels){\
    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 69 70 71 72 73
    }else if(channels==6){\
        ff_float_to_int16_interleave6_##cpu(dst, src, len);\
    }else\
        float_to_int16_interleave_misc_##cpu(dst, src, len, channels);\
}

74 75 76
FLOAT_TO_INT16_INTERLEAVE(3dnow)
FLOAT_TO_INT16_INTERLEAVE(sse)
FLOAT_TO_INT16_INTERLEAVE(sse2)
77

78 79 80
static void float_to_int16_interleave_3dnowext(int16_t *dst, const float **src,
                                               long len, int channels)
{
81
    if(channels==6)
82
        ff_float_to_int16_interleave6_3dnowext(dst, src, len);
83 84 85 86
    else
        float_to_int16_interleave_3dnow(dst, src, len, channels);
}

87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113
void ff_float_interleave2_mmx(float *dst, const float **src, unsigned int len);
void ff_float_interleave2_sse(float *dst, const float **src, unsigned int len);

void ff_float_interleave6_mmx(float *dst, const float **src, unsigned int len);
void ff_float_interleave6_sse(float *dst, const float **src, unsigned int len);

static void float_interleave_mmx(float *dst, const float **src,
                                 unsigned int len, int channels)
{
    if (channels == 2) {
        ff_float_interleave2_mmx(dst, src, len);
    } else if (channels == 6)
        ff_float_interleave6_mmx(dst, src, len);
    else
        ff_float_interleave_c(dst, src, len, channels);
}

static void float_interleave_sse(float *dst, const float **src,
                                 unsigned int len, int channels)
{
    if (channels == 2) {
        ff_float_interleave2_sse(dst, src, len);
    } else if (channels == 6)
        ff_float_interleave6_sse(dst, src, len);
    else
        ff_float_interleave_c(dst, src, len, channels);
}
114
#endif /* HAVE_YASM */
115

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;
123 124 125 126 127
    }
    if (EXTERNAL_AMD3DNOW(cpu_flags)) {
        if (!(avctx->flags & CODEC_FLAG_BITEXACT)) {
            c->float_to_int16            = ff_float_to_int16_3dnow;
            c->float_to_int16_interleave = float_to_int16_interleave_3dnow;
128
        }
129 130 131 132
    }
    if (EXTERNAL_AMD3DNOWEXT(cpu_flags)) {
        if (!(avctx->flags & CODEC_FLAG_BITEXACT)) {
            c->float_to_int16_interleave = float_to_int16_interleave_3dnowext;
133 134
        }
    }
135 136 137 138 139 140 141 142 143 144 145
    if (EXTERNAL_SSE(cpu_flags)) {
        c->int32_to_float_fmul_scalar = ff_int32_to_float_fmul_scalar_sse;
        c->float_to_int16             = ff_float_to_int16_sse;
        c->float_to_int16_interleave  = float_to_int16_interleave_sse;
        c->float_interleave           = float_interleave_sse;
    }
    if (EXTERNAL_SSE2(cpu_flags)) {
        c->int32_to_float_fmul_scalar = ff_int32_to_float_fmul_scalar_sse2;
        c->float_to_int16             = ff_float_to_int16_sse2;
        c->float_to_int16_interleave  = float_to_int16_interleave_sse2;
    }
146
#endif /* HAVE_YASM */
147
}