Statistics
| Branch: | Revision:

ffmpeg / libavutil / x86 / cpu.c @ b7195837

History | View | Annotate | Download (4.8 KB)

1
/*
2
 * CPU detection code, extracted from mmx.h
3
 * (c)1997-99 by H. Dietz and R. Fisher
4
 * Converted to C and improved by Fabrice Bellard.
5
 *
6
 * This file is part of FFmpeg.
7
 *
8
 * FFmpeg 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.
12
 *
13
 * FFmpeg 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.
17
 *
18
 * You should have received a copy of the GNU Lesser General Public
19
 * License along with FFmpeg; if not, write to the Free Software
20
 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
21
 */
22

    
23
#include <stdlib.h>
24
#include <string.h>
25
#include "libavutil/x86_cpu.h"
26
#include "libavutil/cpu.h"
27

    
28
/* ebx saving is necessary for PIC. gcc seems unable to see it alone */
29
#define cpuid(index,eax,ebx,ecx,edx)\
30
    __asm__ volatile\
31
        ("mov %%"REG_b", %%"REG_S"\n\t"\
32
         "cpuid\n\t"\
33
         "xchg %%"REG_b", %%"REG_S\
34
         : "=a" (eax), "=S" (ebx),\
35
           "=c" (ecx), "=d" (edx)\
36
         : "0" (index));
37

    
38
/* Function to test if multimedia instructions are supported...  */
39
int ff_get_cpu_flags_x86(void)
40
{
41
    int rval = 0;
42
    int eax, ebx, ecx, edx;
43
    int max_std_level, max_ext_level, std_caps=0, ext_caps=0;
44
    int family=0, model=0;
45
    union { int i[3]; char c[12]; } vendor;
46

    
47
#if ARCH_X86_32
48
    x86_reg a, c;
49
    __asm__ volatile (
50
        /* See if CPUID instruction is supported ... */
51
        /* ... Get copies of EFLAGS into eax and ecx */
52
        "pushfl\n\t"
53
        "pop %0\n\t"
54
        "mov %0, %1\n\t"
55

    
56
        /* ... Toggle the ID bit in one copy and store */
57
        /*     to the EFLAGS reg */
58
        "xor $0x200000, %0\n\t"
59
        "push %0\n\t"
60
        "popfl\n\t"
61

    
62
        /* ... Get the (hopefully modified) EFLAGS */
63
        "pushfl\n\t"
64
        "pop %0\n\t"
65
        : "=a" (a), "=c" (c)
66
        :
67
        : "cc"
68
        );
69

    
70
    if (a == c)
71
        return 0; /* CPUID not supported */
72
#endif
73

    
74
    cpuid(0, max_std_level, vendor.i[0], vendor.i[2], vendor.i[1]);
75

    
76
    if(max_std_level >= 1){
77
        cpuid(1, eax, ebx, ecx, std_caps);
78
        family = ((eax>>8)&0xf) + ((eax>>20)&0xff);
79
        model  = ((eax>>4)&0xf) + ((eax>>12)&0xf0);
80
        if (std_caps & (1<<23))
81
            rval |= AV_CPU_FLAG_MMX;
82
        if (std_caps & (1<<25))
83
            rval |= AV_CPU_FLAG_MMX2
84
#if HAVE_SSE
85
                  | AV_CPU_FLAG_SSE;
86
        if (std_caps & (1<<26))
87
            rval |= AV_CPU_FLAG_SSE2;
88
        if (ecx & 1)
89
            rval |= AV_CPU_FLAG_SSE3;
90
        if (ecx & 0x00000200 )
91
            rval |= AV_CPU_FLAG_SSSE3;
92
        if (ecx & 0x00080000 )
93
            rval |= AV_CPU_FLAG_SSE4;
94
        if (ecx & 0x00100000 )
95
            rval |= AV_CPU_FLAG_SSE42;
96
#endif
97
                  ;
98
    }
99

    
100
    cpuid(0x80000000, max_ext_level, ebx, ecx, edx);
101

    
102
    if(max_ext_level >= 0x80000001){
103
        cpuid(0x80000001, eax, ebx, ecx, ext_caps);
104
        if (ext_caps & (1<<31))
105
            rval |= AV_CPU_FLAG_3DNOW;
106
        if (ext_caps & (1<<30))
107
            rval |= AV_CPU_FLAG_3DNOWEXT;
108
        if (ext_caps & (1<<23))
109
            rval |= AV_CPU_FLAG_MMX;
110
        if (ext_caps & (1<<22))
111
            rval |= AV_CPU_FLAG_MMX2;
112

    
113
        /* Allow for selectively disabling SSE2 functions on AMD processors
114
           with SSE2 support but not SSE4a. This includes Athlon64, some
115
           Opteron, and some Sempron processors. MMX, SSE, or 3DNow! are faster
116
           than SSE2 often enough to utilize this special-case flag.
117
           AV_CPU_FLAG_SSE2 and AV_CPU_FLAG_SSE2SLOW are both set in this case
118
           so that SSE2 is used unless explicitly disabled by checking
119
           AV_CPU_FLAG_SSE2SLOW. */
120
        if (!strncmp(vendor.c, "AuthenticAMD", 12) &&
121
            rval & AV_CPU_FLAG_SSE2 && !(ecx & 0x00000040)) {
122
            rval |= AV_CPU_FLAG_SSE2SLOW;
123
        }
124
    }
125

    
126
    if (!strncmp(vendor.c, "GenuineIntel", 12) &&
127
        family == 6 && (model == 9 || model == 13 || model == 14)) {
128
        /* 6/9 (pentium-m "banias"), 6/13 (pentium-m "dothan"), and 6/14 (core1 "yonah")
129
         * theoretically support sse2, but it's usually slower than mmx,
130
         * so let's just pretend they don't. AV_CPU_FLAG_SSE2 is disabled and
131
         * AV_CPU_FLAG_SSE2SLOW is enabled so that SSE2 is not used unless
132
         * explicitly enabled by checking AV_CPU_FLAG_SSE2SLOW. The same
133
         * situation applies for AV_CPU_FLAG_SSE3 and AV_CPU_FLAG_SSE3SLOW. */
134
        if (rval & AV_CPU_FLAG_SSE2) rval ^= AV_CPU_FLAG_SSE2SLOW|AV_CPU_FLAG_SSE2;
135
        if (rval & AV_CPU_FLAG_SSE3) rval ^= AV_CPU_FLAG_SSE3SLOW|AV_CPU_FLAG_SSE3;
136
    }
137

    
138
    return rval;
139
}