research.m1stereo.tv/gitweb/ - melted/blob - src/modules/xine/cpu_accel.c

   1 /*
   2  * cpu_accel.c
   3  * Copyright (C) 1999-2001 Aaron Holtzman <aholtzma@ess.engr.uvic.ca>
   4  *
   5  * This file is part of mpeg2dec, a free MPEG-2 video stream decoder.
   6  *
   7  * mpeg2dec is free software; you can redistribute it and/or modify
   8  * it under the terms of the GNU General Public License as published by
   9  * the Free Software Foundation; either version 2 of the License, or
  10  * (at your option) any later version.
  11  *
  12  * mpeg2dec 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
  15  * GNU General Public License for more details.
  16  *
  17  * You should have received a copy of the GNU General Public License
  18  * along with this program; if not, write to the Free Software
  19  * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
  20  */
  21
  22 //#include "config.h"
  23
  24 #include <stdio.h>
  25 #include <stdlib.h>
  26 #include <inttypes.h>
  27 #include <signal.h>
  28 #include <setjmp.h>
  29 #include <dlfcn.h>
  30
  31 #define LOG_MODULE "cpu_accel"
  32 #define LOG_VERBOSE
  33 /*
  34 #define LOG
  35 */
  36
  37 #include "xineutils.h"
  38
  39 #if defined(ARCH_X86) || defined(ARCH_X86_64)
  40 #if defined __x86_64__
  41 static uint32_t arch_accel (void)
  42 {
  43   uint32_t caps;
  44   /* No need to test for this on AMD64, we know what the
  45      platform has.  */
  46   caps = MM_ACCEL_X86_MMX | MM_ACCEL_X86_SSE | MM_ACCEL_X86_MMXEXT | MM_ACCEL_X86_SSE2;
  47
  48   return caps;
  49 }
  50 #else
  51 static uint32_t arch_accel (void)
  52 {
  53 #ifndef _MSC_VER
  54
  55   uint32_t eax, ebx, ecx, edx;
  56   int AMD;
  57   uint32_t caps;
  58
  59 #ifndef PIC
  60 #define cpuid(op,eax,ebx,ecx,edx)       \
  61     __asm__ ("cpuid"                        \
  62          : "=a" (eax),                  \
  63            "=b" (ebx),                  \
  64            "=c" (ecx),                  \
  65            "=d" (edx)                   \
  66          : "a" (op)                     \
  67          : "cc")
  68 #else   /* PIC version : save ebx */
  69 #define cpuid(op,eax,ebx,ecx,edx)       \
  70     __asm__ ("pushl %%ebx\n\t"              \
  71          "cpuid\n\t"                    \
  72          "movl %%ebx,%1\n\t"            \
  73          "popl %%ebx"                   \
  74          : "=a" (eax),                  \
  75            "=r" (ebx),                  \
  76            "=c" (ecx),                  \
  77            "=d" (edx)                   \
  78          : "a" (op)                     \
  79          : "cc")
  80 #endif
  81
  82   __asm__ ("pushfl\n\t"
  83        "pushfl\n\t"
  84        "popl %0\n\t"
  85        "movl %0,%1\n\t"
  86        "xorl $0x200000,%0\n\t"
  87        "pushl %0\n\t"
  88        "popfl\n\t"
  89        "pushfl\n\t"
  90        "popl %0\n\t"
  91        "popfl"
  92        : "=r" (eax),
  93        "=r" (ebx)
  94        :
  95        : "cc");
  96
  97   if (eax == ebx)             /* no cpuid */
  98     return 0;
  99
 100   cpuid (0x00000000, eax, ebx, ecx, edx);
 101   if (!eax)                   /* vendor string only */
 102     return 0;
 103
 104   AMD = (ebx == 0x68747541) && (ecx == 0x444d4163) && (edx == 0x69746e65);
 105
 106   cpuid (0x00000001, eax, ebx, ecx, edx);
 107   if (! (edx & 0x00800000))   /* no MMX */
 108     return 0;
 109
 110   caps = MM_ACCEL_X86_MMX;
 111   if (edx & 0x02000000)       /* SSE - identical to AMD MMX extensions */
 112     caps |= MM_ACCEL_X86_SSE | MM_ACCEL_X86_MMXEXT;
 113
 114   if (edx & 0x04000000)       /* SSE2 */
 115     caps |= MM_ACCEL_X86_SSE2;
 116
 117   cpuid (0x80000000, eax, ebx, ecx, edx);
 118   if (eax < 0x80000001)       /* no extended capabilities */
 119     return caps;
 120
 121   cpuid (0x80000001, eax, ebx, ecx, edx);
 122
 123   if (edx & 0x80000000)
 124     caps |= MM_ACCEL_X86_3DNOW;
 125
 126   if (AMD && (edx & 0x00400000))      /* AMD MMX extensions */
 127     caps |= MM_ACCEL_X86_MMXEXT;
 128
 129   return caps;
 130 #else /* _MSC_VER */
 131   return 0;
 132 #endif
 133 }
 134 #endif /* x86_64 */
 135
 136 static jmp_buf sigill_return;
 137
 138 static void sigill_handler (int n) {
 139   longjmp(sigill_return, 1);
 140 }
 141 #endif /* ARCH_X86 */
 142
 143 #if defined (ARCH_PPC) && defined (ENABLE_ALTIVEC)
 144 static sigjmp_buf jmpbuf;
 145 static volatile sig_atomic_t canjump = 0;
 146
 147 static void sigill_handler (int sig)
 148 {
 149     if (!canjump) {
 150         signal (sig, SIG_DFL);
 151         raise (sig);
 152     }
 153
 154     canjump = 0;
 155     siglongjmp (jmpbuf, 1);
 156 }
 157
 158 static uint32_t arch_accel (void)
 159 {
 160     signal (SIGILL, sigill_handler);
 161     if (sigsetjmp (jmpbuf, 1)) {
 162         signal (SIGILL, SIG_DFL);
 163         return 0;
 164     }
 165
 166     canjump = 1;
 167
 168     __asm__ volatile ("mtspr 256, %0\n\t"
 169                   "vand %%v0, %%v0, %%v0"
 170                   :
 171                   : "r" (-1));
 172
 173     signal (SIGILL, SIG_DFL);
 174     return MM_ACCEL_PPC_ALTIVEC;
 175 }
 176 #endif /* ARCH_PPC */
 177
 178 uint32_t xine_mm_accel (void)
 179 {
 180   static int initialized = 0;
 181   static uint32_t accel;
 182
 183   if (!initialized) {
 184 #if defined (ARCH_X86) || (defined (ARCH_PPC) && defined (ENABLE_ALTIVEC))
 185     accel = arch_accel ();
 186 #elif defined (HAVE_MLIB)
 187 #ifdef MLIB_LAZYLOAD
 188     void *hndl;
 189
 190     if ((hndl = dlopen("libmlib.so.2", RTLD_LAZY | RTLD_GLOBAL | RTLD_NODELETE)) == NULL) {
 191       accel = 0;
 192     }
 193     else {
 194       dlclose(hndl);
 195       accel = MM_ACCEL_MLIB;
 196     }
 197 #else
 198     accel = MM_ACCEL_MLIB;
 199 #endif
 200 #else
 201     accel = 0;
 202 #endif
 203
 204 #if defined(ARCH_X86) || defined(ARCH_X86_64)
 205 #ifndef _MSC_VER
 206     /* test OS support for SSE */
 207     if( accel & MM_ACCEL_X86_SSE ) {
 208       void (*old_sigill_handler)(int);
 209
 210       old_sigill_handler = signal (SIGILL, sigill_handler);
 211
 212       if (setjmp(sigill_return)) {
 213         lprintf ("OS doesn't support SSE instructions.\n");
 214         accel &= ~(MM_ACCEL_X86_SSE|MM_ACCEL_X86_SSE2);
 215       } else {
 216         __asm__ volatile ("xorps %xmm0, %xmm0");
 217       }
 218
 219       signal (SIGILL, old_sigill_handler);
 220     }
 221 #endif /* _MSC_VER */
 222 #endif /* ARCH_X86 || ARCH_X86_64 */
 223
 224     if(getenv("XINE_NO_ACCEL")) {
 225       accel = 0;
 226     }
 227
 228     initialized = 1;
 229   }
 230
 231   return accel;
 232 }