Statistics
| Branch: | Revision:

ffmpeg / libavcodec / x86 / mpegaudiodec_mmx.c @ 2912e87a

History | View | Annotate | Download (4.99 KB)

1
/*
2
 * MMX optimized MP3 decoding functions
3
 * Copyright (c) 2010 Vitor Sessak
4
 *
5
 * This file is part of Libav.
6
 *
7
 * Libav is free software; you can redistribute it and/or
8
 * modify it under the terms of the GNU Lesser General Public
9
 * License as published by the Free Software Foundation; either
10
 * version 2.1 of the License, or (at your option) any later version.
11
 *
12
 * Libav is distributed in the hope that it will be useful,
13
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
15
 * Lesser General Public License for more details.
16
 *
17
 * You should have received a copy of the GNU Lesser General Public
18
 * License along with Libav; if not, write to the Free Software
19
 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
20
 */
21

    
22
#include "libavutil/cpu.h"
23
#include "libavutil/x86_cpu.h"
24

    
25
#define CONFIG_FLOAT 1
26
#include "libavcodec/mpegaudio.h"
27

    
28
#define MACS(rt, ra, rb) rt+=(ra)*(rb)
29
#define MLSS(rt, ra, rb) rt-=(ra)*(rb)
30

    
31
#define SUM8(op, sum, w, p)               \
32
{                                         \
33
    op(sum, (w)[0 * 64], (p)[0 * 64]);    \
34
    op(sum, (w)[1 * 64], (p)[1 * 64]);    \
35
    op(sum, (w)[2 * 64], (p)[2 * 64]);    \
36
    op(sum, (w)[3 * 64], (p)[3 * 64]);    \
37
    op(sum, (w)[4 * 64], (p)[4 * 64]);    \
38
    op(sum, (w)[5 * 64], (p)[5 * 64]);    \
39
    op(sum, (w)[6 * 64], (p)[6 * 64]);    \
40
    op(sum, (w)[7 * 64], (p)[7 * 64]);    \
41
}
42

    
43
static void apply_window(const float *buf, const float *win1,
44
                         const float *win2, float *sum1, float *sum2, int len)
45
{
46
    x86_reg count = - 4*len;
47
    const float *win1a = win1+len;
48
    const float *win2a = win2+len;
49
    const float *bufa  = buf+len;
50
    float *sum1a = sum1+len;
51
    float *sum2a = sum2+len;
52

    
53

    
54
#define MULT(a, b)                                 \
55
    "movaps " #a "(%1,%0), %%xmm1           \n\t"  \
56
    "movaps " #a "(%3,%0), %%xmm2           \n\t"  \
57
    "mulps         %%xmm2, %%xmm1           \n\t"  \
58
    "subps         %%xmm1, %%xmm0           \n\t"  \
59
    "mulps  " #b "(%2,%0), %%xmm2           \n\t"  \
60
    "subps         %%xmm2, %%xmm4           \n\t"  \
61

    
62
    __asm__ volatile(
63
            "1:                                   \n\t"
64
            "xorps       %%xmm0, %%xmm0           \n\t"
65
            "xorps       %%xmm4, %%xmm4           \n\t"
66

    
67
            MULT(   0,   0)
68
            MULT( 256,  64)
69
            MULT( 512, 128)
70
            MULT( 768, 192)
71
            MULT(1024, 256)
72
            MULT(1280, 320)
73
            MULT(1536, 384)
74
            MULT(1792, 448)
75

    
76
            "movaps      %%xmm0, (%4,%0)          \n\t"
77
            "movaps      %%xmm4, (%5,%0)          \n\t"
78
            "add            $16,  %0              \n\t"
79
            "jl              1b                   \n\t"
80
            :"+&r"(count)
81
            :"r"(win1a), "r"(win2a), "r"(bufa), "r"(sum1a), "r"(sum2a)
82
            );
83

    
84
#undef MULT
85
}
86

    
87
static void apply_window_mp3(float *in, float *win, int *unused, float *out,
88
                             int incr)
89
{
90
    LOCAL_ALIGNED_16(float, suma, [17]);
91
    LOCAL_ALIGNED_16(float, sumb, [17]);
92
    LOCAL_ALIGNED_16(float, sumc, [17]);
93
    LOCAL_ALIGNED_16(float, sumd, [17]);
94

    
95
    float sum;
96

    
97
    /* copy to avoid wrap */
98
    memcpy(in + 512, in, 32 * sizeof(*in));
99

    
100
    apply_window(in + 16, win     , win + 512, suma, sumc, 16);
101
    apply_window(in + 32, win + 48, win + 640, sumb, sumd, 16);
102

    
103
    SUM8(MACS, suma[0], win + 32, in + 48);
104

    
105
    sumc[ 0] = 0;
106
    sumb[16] = 0;
107
    sumd[16] = 0;
108

    
109
#define SUMS(suma, sumb, sumc, sumd, out1, out2)               \
110
            "movups " #sumd "(%4),       %%xmm0          \n\t" \
111
            "shufps         $0x1b,       %%xmm0, %%xmm0  \n\t" \
112
            "subps  " #suma "(%1),       %%xmm0          \n\t" \
113
            "movaps        %%xmm0," #out1 "(%0)          \n\t" \
114
\
115
            "movups " #sumc "(%3),       %%xmm0          \n\t" \
116
            "shufps         $0x1b,       %%xmm0, %%xmm0  \n\t" \
117
            "addps  " #sumb "(%2),       %%xmm0          \n\t" \
118
            "movaps        %%xmm0," #out2 "(%0)          \n\t"
119

    
120
    if (incr == 1) {
121
        __asm__ volatile(
122
            SUMS( 0, 48,  4, 52,  0, 112)
123
            SUMS(16, 32, 20, 36, 16,  96)
124
            SUMS(32, 16, 36, 20, 32,  80)
125
            SUMS(48,  0, 52,  4, 48,  64)
126

    
127
            :"+&r"(out)
128
            :"r"(&suma[0]), "r"(&sumb[0]), "r"(&sumc[0]), "r"(&sumd[0])
129
            :"memory"
130
            );
131
        out += 16*incr;
132
    } else {
133
        int j;
134
        float *out2 = out + 32 * incr;
135
        out[0  ]  = -suma[   0];
136
        out += incr;
137
        out2 -= incr;
138
        for(j=1;j<16;j++) {
139
            *out  = -suma[   j] + sumd[16-j];
140
            *out2 =  sumb[16-j] + sumc[   j];
141
            out  += incr;
142
            out2 -= incr;
143
        }
144
    }
145

    
146
    sum = 0;
147
    SUM8(MLSS, sum, win + 16 + 32, in + 32);
148
    *out = sum;
149
}
150

    
151
void ff_mpegaudiodec_init_mmx(MPADecodeContext *s)
152
{
153
    int mm_flags = av_get_cpu_flags();
154

    
155
    if (mm_flags & AV_CPU_FLAG_SSE2) {
156
        s->apply_window_mp3 = apply_window_mp3;
157
    }
158
}