cpu.c 9.15 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>
8
 *          Fiona Glaser <fiona@x264.com>
Laurent Aimar's avatar
Laurent Aimar committed
9 10 11 12 13 14 15 16 17 18 19 20 21
 *
 * 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
22
 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02111, USA.
Laurent Aimar's avatar
Laurent Aimar committed
23 24
 *****************************************************************************/

25 26 27 28
#if defined(HAVE_PTHREAD) && defined(SYS_LINUX)
#define _GNU_SOURCE
#include <sched.h>
#endif
29 30 31
#ifdef SYS_BEOS
#include <kernel/OS.h>
#endif
32
#if defined(SYS_MACOSX) || defined(SYS_FREEBSD)
33 34 35
#include <sys/types.h>
#include <sys/sysctl.h>
#endif
36

Loren Merritt's avatar
Loren Merritt committed
37
#include "common.h"
38
#include "cpu.h"
Laurent Aimar's avatar
Laurent Aimar committed
39

40 41 42
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
43 44
    {"MMX2",    X264_CPU_MMX|X264_CPU_MMXEXT},
    {"MMXEXT",  X264_CPU_MMX|X264_CPU_MMXEXT},
45 46
//  {"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},
47
    {"SSE2",    X264_CPU_MMX|X264_CPU_MMXEXT|X264_CPU_SSE|X264_CPU_SSE2},
48
    {"SSE2Fast",X264_CPU_MMX|X264_CPU_MMXEXT|X264_CPU_SSE|X264_CPU_SSE2|X264_CPU_SSE2_IS_FAST},
49 50
    {"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},
51
    {"PHADD",   X264_CPU_MMX|X264_CPU_MMXEXT|X264_CPU_SSE|X264_CPU_SSE2|X264_CPU_SSE3|X264_CPU_SSSE3|X264_CPU_PHADD_IS_FAST},
52 53
    {"SSE4.1",  X264_CPU_MMX|X264_CPU_MMXEXT|X264_CPU_SSE|X264_CPU_SSE2|X264_CPU_SSE3|X264_CPU_SSSE3|X264_CPU_SSE4},
    {"SSE4.2",  X264_CPU_MMX|X264_CPU_MMXEXT|X264_CPU_SSE|X264_CPU_SSE2|X264_CPU_SSE3|X264_CPU_SSSE3|X264_CPU_SSE4|X264_CPU_SSE42},
54 55
    {"Cache32", X264_CPU_CACHELINE_32},
    {"Cache64", X264_CPU_CACHELINE_64},
Fiona Glaser's avatar
Fiona Glaser committed
56
    {"SSEMisalign", X264_CPU_SSE_MISALIGN},
57
    {"Slow_mod4_stack", X264_CPU_STACK_MOD4},
58 59 60
    {"", 0},
};

Fiona Glaser's avatar
Fiona Glaser committed
61

Loren Merritt's avatar
Loren Merritt committed
62
#ifdef HAVE_MMX
Laurent Aimar's avatar
Laurent Aimar committed
63 64 65 66 67 68 69
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;
70 71 72
    uint32_t vendor[4] = {0};
    int max_extended_cap;
    int cache;
Laurent Aimar's avatar
Laurent Aimar committed
73

74
#ifndef ARCH_X86_64
Laurent Aimar's avatar
Laurent Aimar committed
75 76
    if( !x264_cpu_cpuid_test() )
        return 0;
77
#endif
Laurent Aimar's avatar
Laurent Aimar committed
78

79
    x264_cpu_cpuid( 0, &eax, vendor+0, vendor+2, vendor+1 );
Laurent Aimar's avatar
Laurent Aimar committed
80 81 82 83
    if( eax == 0 )
        return 0;

    x264_cpu_cpuid( 1, &eax, &ebx, &ecx, &edx );
84 85 86
    if( edx&0x00800000 )
        cpu |= X264_CPU_MMX;
    else
Laurent Aimar's avatar
Laurent Aimar committed
87
        return 0;
88
    if( edx&0x02000000 )
Laurent Aimar's avatar
Laurent Aimar committed
89
        cpu |= X264_CPU_MMXEXT|X264_CPU_SSE;
90
    if( edx&0x04000000 )
Laurent Aimar's avatar
Laurent Aimar committed
91
        cpu |= X264_CPU_SSE2;
92
    if( ecx&0x00000001 )
93
        cpu |= X264_CPU_SSE3;
94
    if( ecx&0x00000200 )
95
        cpu |= X264_CPU_SSSE3;
96 97
    if( ecx&0x00080000 )
        cpu |= X264_CPU_SSE4;
98 99
    if( ecx&0x00100000 )
        cpu |= X264_CPU_SSE42;
Laurent Aimar's avatar
Laurent Aimar committed
100

101 102 103 104 105
    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
106
    x264_cpu_cpuid( 0x80000000, &eax, &ebx, &ecx, &edx );
107 108 109
    max_extended_cap = eax;

    if( !strcmp((char*)vendor, "AuthenticAMD") && max_extended_cap >= 0x80000001 )
Laurent Aimar's avatar
Laurent Aimar committed
110
    {
111 112 113
        x264_cpu_cpuid( 0x80000001, &eax, &ebx, &ecx, &edx );
        if( edx&0x00400000 )
            cpu |= X264_CPU_MMXEXT;
114 115 116
        if( cpu & X264_CPU_SSE2 )
        {
            if( ecx&0x00000040 ) /* SSE4a */
Fiona Glaser's avatar
Fiona Glaser committed
117
            {
118
                cpu |= X264_CPU_SSE2_IS_FAST;
Fiona Glaser's avatar
Fiona Glaser committed
119 120 121
                cpu |= X264_CPU_SSE_MISALIGN;
                x264_cpu_mask_misalign_sse();
            }
122 123 124
            else
                cpu |= X264_CPU_SSE2_IS_SLOW;
        }
Laurent Aimar's avatar
Laurent Aimar committed
125 126
    }

127
    if( !strcmp((char*)vendor, "GenuineIntel") )
Laurent Aimar's avatar
Laurent Aimar committed
128
    {
129 130 131 132 133 134 135 136 137 138 139 140 141
        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
142
    }
143

144
    if( (!strcmp((char*)vendor, "GenuineIntel") || !strcmp((char*)vendor, "CyrixInstead")) && !(cpu&X264_CPU_SSE42))
Laurent Aimar's avatar
Laurent Aimar committed
145
    {
146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176
        /* 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
177

178 179 180 181 182 183 184
        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" );
    }
185

186 187 188 189
#ifdef BROKEN_STACK_ALIGNMENT
    cpu |= X264_CPU_STACK_MOD4;
#endif

Laurent Aimar's avatar
Laurent Aimar committed
190 191 192
    return cpu;
}

193
#elif defined( ARCH_PPC )
Laurent Aimar's avatar
Laurent Aimar committed
194

195 196
#ifdef SYS_MACOSX
#include <sys/sysctl.h>
Laurent Aimar's avatar
Laurent Aimar committed
197 198
uint32_t x264_cpu_detect( void )
{
199
    /* Thank you VLC */
Laurent Aimar's avatar
Laurent Aimar committed
200 201 202 203 204 205 206 207 208 209 210 211 212 213
    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;
}

214
#elif defined( SYS_LINUX )
215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231
#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 );
}

232 233
uint32_t x264_cpu_detect( void )
{
234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251
    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 );

252 253 254 255
    return X264_CPU_ALTIVEC;
}
#endif

Laurent Aimar's avatar
Laurent Aimar committed
256 257 258 259 260 261 262
#else

uint32_t x264_cpu_detect( void )
{
    return 0;
}

Loren Merritt's avatar
Loren Merritt committed
263 264 265 266
#endif

#ifndef HAVE_MMX
void x264_emms( void )
Laurent Aimar's avatar
Laurent Aimar committed
267 268 269
{
}
#endif
270 271 272 273


int x264_cpu_num_processors( void )
{
274 275 276
#if !defined(HAVE_PTHREAD)
    return 1;

277
#elif defined(_WIN32)
Loren Merritt's avatar
Loren Merritt committed
278 279 280
    return pthread_num_processors_np();

#elif defined(SYS_LINUX)
281
    unsigned int bit;
282
    int np;
283 284 285 286 287
    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;
288 289
    return np;

290 291 292 293 294
#elif defined(SYS_BEOS)
    system_info info;
    get_system_info( &info );
    return info.cpu_count;

295
#elif defined(SYS_MACOSX) || defined(SYS_FREEBSD)
296 297 298 299 300
    int numberOfCPUs;
    size_t length = sizeof( numberOfCPUs );
    if( sysctlbyname("hw.ncpu", &numberOfCPUs, &length, NULL, 0) )
    {
        numberOfCPUs = 1;
301
    }
302
    return numberOfCPUs;
303

304
#else
305 306
    return 1;
#endif
307
}