cpu.c 8.74 KB
Newer Older
Laurent Aimar's avatar
Laurent Aimar committed
1 2 3
/*****************************************************************************
 * cpu.c: h264 encoder library
 *****************************************************************************
4
 * Copyright (C) 2003-2008 x264 project
Laurent Aimar's avatar
Laurent Aimar committed
5
 *
6 7
 * Authors: Loren Merritt <lorenm@u.washington.edu>
 *          Laurent Aimar <fenrir@via.ecp.fr>
Laurent Aimar's avatar
Laurent Aimar committed
8 9 10 11 12 13 14 15 16 17 18 19 20
 *
 * This program is free software; you can redistribute it and/or modify
 * it under the terms of the GNU General Public License as published by
 * the Free Software Foundation; either version 2 of the License, or
 * (at your option) any later version.
 *
 * This program is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 * GNU General Public License for more details.
 *
 * You should have received a copy of the GNU General Public License
 * along with this program; if not, write to the Free Software
21
 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02111, USA.
Laurent Aimar's avatar
Laurent Aimar committed
22 23
 *****************************************************************************/

24 25 26 27
#if defined(HAVE_PTHREAD) && defined(SYS_LINUX)
#define _GNU_SOURCE
#include <sched.h>
#endif
28 29 30
#ifdef SYS_BEOS
#include <kernel/OS.h>
#endif
31
#if defined(SYS_MACOSX) || defined(SYS_FREEBSD)
32 33 34
#include <sys/types.h>
#include <sys/sysctl.h>
#endif
35

Loren Merritt's avatar
Loren Merritt committed
36
#include "common.h"
37
#include "cpu.h"
Laurent Aimar's avatar
Laurent Aimar committed
38

39 40 41
const x264_cpu_name_t x264_cpu_names[] = {
    {"Altivec", X264_CPU_ALTIVEC},
//  {"MMX",     X264_CPU_MMX}, // we don't support asm on mmx1 cpus anymore
42 43
    {"MMX2",    X264_CPU_MMX|X264_CPU_MMXEXT},
    {"MMXEXT",  X264_CPU_MMX|X264_CPU_MMXEXT},
44 45
//  {"SSE",     X264_CPU_MMX|X264_CPU_MMXEXT|X264_CPU_SSE}, // there are no sse1 functions in x264
    {"SSE2Slow",X264_CPU_MMX|X264_CPU_MMXEXT|X264_CPU_SSE|X264_CPU_SSE2|X264_CPU_SSE2_IS_SLOW},
46
    {"SSE2",    X264_CPU_MMX|X264_CPU_MMXEXT|X264_CPU_SSE|X264_CPU_SSE2},
47
    {"SSE2Fast",X264_CPU_MMX|X264_CPU_MMXEXT|X264_CPU_SSE|X264_CPU_SSE2|X264_CPU_SSE2_IS_FAST},
48 49
    {"SSE3",    X264_CPU_MMX|X264_CPU_MMXEXT|X264_CPU_SSE|X264_CPU_SSE2|X264_CPU_SSE3},
    {"SSSE3",   X264_CPU_MMX|X264_CPU_MMXEXT|X264_CPU_SSE|X264_CPU_SSE2|X264_CPU_SSE3|X264_CPU_SSSE3},
50
    {"PHADD",   X264_CPU_MMX|X264_CPU_MMXEXT|X264_CPU_SSE|X264_CPU_SSE2|X264_CPU_SSE3|X264_CPU_SSSE3|X264_CPU_PHADD_IS_FAST},
51
    {"SSE4",    X264_CPU_MMX|X264_CPU_MMXEXT|X264_CPU_SSE|X264_CPU_SSE2|X264_CPU_SSE3|X264_CPU_SSSE3|X264_CPU_SSE4},
52 53
    {"Cache32", X264_CPU_CACHELINE_32},
    {"Cache64", X264_CPU_CACHELINE_64},
54
    {"Slow_mod4_stack", X264_CPU_STACK_MOD4},
55 56 57
    {"", 0},
};

Loren Merritt's avatar
Loren Merritt committed
58
#ifdef HAVE_MMX
Laurent Aimar's avatar
Laurent Aimar committed
59 60 61 62 63 64 65
extern int  x264_cpu_cpuid_test( void );
extern uint32_t  x264_cpu_cpuid( uint32_t op, uint32_t *eax, uint32_t *ebx, uint32_t *ecx, uint32_t *edx );

uint32_t x264_cpu_detect( void )
{
    uint32_t cpu = 0;
    uint32_t eax, ebx, ecx, edx;
66 67 68
    uint32_t vendor[4] = {0};
    int max_extended_cap;
    int cache;
Laurent Aimar's avatar
Laurent Aimar committed
69

70
#ifndef ARCH_X86_64
Laurent Aimar's avatar
Laurent Aimar committed
71 72
    if( !x264_cpu_cpuid_test() )
        return 0;
73
#endif
Laurent Aimar's avatar
Laurent Aimar committed
74

75
    x264_cpu_cpuid( 0, &eax, vendor+0, vendor+2, vendor+1 );
Laurent Aimar's avatar
Laurent Aimar committed
76 77 78 79
    if( eax == 0 )
        return 0;

    x264_cpu_cpuid( 1, &eax, &ebx, &ecx, &edx );
80 81 82
    if( edx&0x00800000 )
        cpu |= X264_CPU_MMX;
    else
Laurent Aimar's avatar
Laurent Aimar committed
83
        return 0;
84
    if( edx&0x02000000 )
Laurent Aimar's avatar
Laurent Aimar committed
85
        cpu |= X264_CPU_MMXEXT|X264_CPU_SSE;
86
    if( edx&0x04000000 )
Laurent Aimar's avatar
Laurent Aimar committed
87
        cpu |= X264_CPU_SSE2;
88
    if( ecx&0x00000001 )
89
        cpu |= X264_CPU_SSE3;
90
    if( ecx&0x00000200 )
91
        cpu |= X264_CPU_SSSE3;
92 93
    if( ecx&0x00080000 )
        cpu |= X264_CPU_SSE4;
Laurent Aimar's avatar
Laurent Aimar committed
94

95 96 97 98 99
    if( cpu & X264_CPU_SSSE3 )
        cpu |= X264_CPU_SSE2_IS_FAST;
    if( cpu & X264_CPU_SSE4 )
        cpu |= X264_CPU_PHADD_IS_FAST;

Laurent Aimar's avatar
Laurent Aimar committed
100
    x264_cpu_cpuid( 0x80000000, &eax, &ebx, &ecx, &edx );
101 102 103
    max_extended_cap = eax;

    if( !strcmp((char*)vendor, "AuthenticAMD") && max_extended_cap >= 0x80000001 )
Laurent Aimar's avatar
Laurent Aimar committed
104
    {
105 106 107
        x264_cpu_cpuid( 0x80000001, &eax, &ebx, &ecx, &edx );
        if( edx&0x00400000 )
            cpu |= X264_CPU_MMXEXT;
108 109 110 111 112 113 114
        if( cpu & X264_CPU_SSE2 )
        {
            if( ecx&0x00000040 ) /* SSE4a */
                cpu |= X264_CPU_SSE2_IS_FAST;
            else
                cpu |= X264_CPU_SSE2_IS_SLOW;
        }
Laurent Aimar's avatar
Laurent Aimar committed
115 116
    }

117
    if( !strcmp((char*)vendor, "GenuineIntel") )
Laurent Aimar's avatar
Laurent Aimar committed
118
    {
119 120 121 122 123 124 125 126 127 128 129 130 131
        int family, model, stepping;
        x264_cpu_cpuid( 1, &eax, &ebx, &ecx, &edx );
        family = ((eax>>8)&0xf) + ((eax>>20)&0xff);
        model  = ((eax>>4)&0xf) + ((eax>>12)&0xf0);
        stepping = eax&0xf;
        /* 6/9 (pentium-m "banias"), 6/13 (pentium-m "dothan"), and 6/14 (core1 "yonah")
         * theoretically support sse2, but it's significantly slower than mmx for
         * almost all of x264's functions, so let's just pretend they don't. */
        if( family==6 && (model==9 || model==13 || model==14) )
        {
            cpu &= ~(X264_CPU_SSE2|X264_CPU_SSE3);
            assert(!(cpu&(X264_CPU_SSSE3|X264_CPU_SSE4)));
        }
Laurent Aimar's avatar
Laurent Aimar committed
132
    }
133 134

    if( !strcmp((char*)vendor, "GenuineIntel") || !strcmp((char*)vendor, "CyrixInstead") )
Laurent Aimar's avatar
Laurent Aimar committed
135
    {
136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166
        /* cacheline size is specified in 3 places, any of which may be missing */
        x264_cpu_cpuid( 1, &eax, &ebx, &ecx, &edx );
        cache = (ebx&0xff00)>>5; // cflush size
        if( !cache && max_extended_cap >= 0x80000006 )
        {
            x264_cpu_cpuid( 0x80000006, &eax, &ebx, &ecx, &edx );
            cache = ecx&0xff; // cacheline size
        }
        if( !cache )
        {
            // Cache and TLB Information
            static const char cache32_ids[] = { 0x0a, 0x0c, 0x41, 0x42, 0x43, 0x44, 0x45, 0x82, 0x83, 0x84, 0x85, 0 };
            static const char cache64_ids[] = { 0x22, 0x23, 0x25, 0x29, 0x2c, 0x46, 0x47, 0x49, 0x60, 0x66, 0x67, 0x68, 0x78, 0x79, 0x7a, 0x7b, 0x7c, 0x7c, 0x7f, 0x86, 0x87, 0 };
            uint32_t buf[4];
            int max, i=0, j;
            do {
                x264_cpu_cpuid( 2, buf+0, buf+1, buf+2, buf+3 );
                max = buf[0]&0xff;
                buf[0] &= ~0xff;
                for(j=0; j<4; j++)
                    if( !(buf[j]>>31) )
                        while( buf[j] )
                        {
                            if( strchr( cache32_ids, buf[j]&0xff ) )
                                cache = 32;
                            if( strchr( cache64_ids, buf[j]&0xff ) )
                                cache = 64;
                            buf[j] >>= 8;
                        }
            } while( ++i < max );
        }
Laurent Aimar's avatar
Laurent Aimar committed
167

168 169 170 171 172 173 174
        if( cache == 32 )
            cpu |= X264_CPU_CACHELINE_32;
        else if( cache == 64 )
            cpu |= X264_CPU_CACHELINE_64;
        else
            fprintf( stderr, "x264 [warning]: unable to determine cacheline size\n" );
    }
175

176 177 178 179
#ifdef BROKEN_STACK_ALIGNMENT
    cpu |= X264_CPU_STACK_MOD4;
#endif

Laurent Aimar's avatar
Laurent Aimar committed
180 181 182
    return cpu;
}

183
#elif defined( ARCH_PPC )
Laurent Aimar's avatar
Laurent Aimar committed
184

185 186
#ifdef SYS_MACOSX
#include <sys/sysctl.h>
Laurent Aimar's avatar
Laurent Aimar committed
187 188
uint32_t x264_cpu_detect( void )
{
189
    /* Thank you VLC */
Laurent Aimar's avatar
Laurent Aimar committed
190 191 192 193 194 195 196 197 198 199 200 201 202 203
    uint32_t cpu = 0;
    int      selectors[2] = { CTL_HW, HW_VECTORUNIT };
    int      has_altivec = 0;
    size_t   length = sizeof( has_altivec );
    int      error = sysctl( selectors, 2, &has_altivec, &length, NULL, 0 );

    if( error == 0 && has_altivec != 0 )
    {
        cpu |= X264_CPU_ALTIVEC;
    }

    return cpu;
}

204
#elif defined( SYS_LINUX )
205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221
#include <signal.h>
#include <setjmp.h>
static sigjmp_buf jmpbuf;
static volatile sig_atomic_t canjump = 0;

static void sigill_handler( int sig )
{
    if( !canjump )
    {
        signal( sig, SIG_DFL );
        raise( sig );
    }

    canjump = 0;
    siglongjmp( jmpbuf, 1 );
}

222 223
uint32_t x264_cpu_detect( void )
{
224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241
    static void (* oldsig)( int );

    oldsig = signal( SIGILL, sigill_handler );
    if( sigsetjmp( jmpbuf, 1 ) )
    {
        signal( SIGILL, oldsig );
        return 0;
    }

    canjump = 1;
    asm volatile( "mtspr 256, %0\n\t"
                  "vand 0, 0, 0\n\t"
                  :
                  : "r"(-1) );
    canjump = 0;

    signal( SIGILL, oldsig );

242 243 244 245
    return X264_CPU_ALTIVEC;
}
#endif

Laurent Aimar's avatar
Laurent Aimar committed
246 247 248 249 250 251 252
#else

uint32_t x264_cpu_detect( void )
{
    return 0;
}

Loren Merritt's avatar
Loren Merritt committed
253 254 255 256
#endif

#ifndef HAVE_MMX
void x264_emms( void )
Laurent Aimar's avatar
Laurent Aimar committed
257 258 259
{
}
#endif
260 261 262 263


int x264_cpu_num_processors( void )
{
264 265 266
#if !defined(HAVE_PTHREAD)
    return 1;

267
#elif defined(_WIN32)
Loren Merritt's avatar
Loren Merritt committed
268 269 270
    return pthread_num_processors_np();

#elif defined(SYS_LINUX)
271
    unsigned int bit;
272
    int np;
273 274 275 276 277
    cpu_set_t p_aff;
    memset( &p_aff, 0, sizeof(p_aff) );
    sched_getaffinity( 0, sizeof(p_aff), &p_aff );
    for( np = 0, bit = 0; bit < sizeof(p_aff); bit++ )
        np += (((uint8_t *)&p_aff)[bit / 8] >> (bit % 8)) & 1;
278 279
    return np;

280 281 282 283 284
#elif defined(SYS_BEOS)
    system_info info;
    get_system_info( &info );
    return info.cpu_count;

285
#elif defined(SYS_MACOSX) || defined(SYS_FREEBSD)
286 287 288 289 290
    int numberOfCPUs;
    size_t length = sizeof( numberOfCPUs );
    if( sysctlbyname("hw.ncpu", &numberOfCPUs, &length, NULL, 0) )
    {
        numberOfCPUs = 1;
291
    }
292
    return numberOfCPUs;
293

294
#else
295 296
    return 1;
#endif
297
}