| OLD | NEW |
| (Empty) |
| 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 "libavutil/x86_cpu.h" | |
| 25 #include "libavcodec/dsputil.h" | |
| 26 | |
| 27 #undef printf | |
| 28 | |
| 29 /* ebx saving is necessary for PIC. gcc seems unable to see it alone */ | |
| 30 #define cpuid(index,eax,ebx,ecx,edx)\ | |
| 31 __asm__ volatile\ | |
| 32 ("mov %%"REG_b", %%"REG_S"\n\t"\ | |
| 33 "cpuid\n\t"\ | |
| 34 "xchg %%"REG_b", %%"REG_S\ | |
| 35 : "=a" (eax), "=S" (ebx),\ | |
| 36 "=c" (ecx), "=d" (edx)\ | |
| 37 : "0" (index)); | |
| 38 | |
| 39 /* Function to test if multimedia instructions are supported... */ | |
| 40 int mm_support(void) | |
| 41 { | |
| 42 int rval = 0; | |
| 43 int eax, ebx, ecx, edx; | |
| 44 int max_std_level, max_ext_level, std_caps=0, ext_caps=0; | |
| 45 int family=0, model=0; | |
| 46 union { int i[3]; char c[12]; } vendor; | |
| 47 | |
| 48 #if ARCH_X86_32 | |
| 49 x86_reg a, c; | |
| 50 __asm__ volatile ( | |
| 51 /* See if CPUID instruction is supported ... */ | |
| 52 /* ... Get copies of EFLAGS into eax and ecx */ | |
| 53 "pushfl\n\t" | |
| 54 "pop %0\n\t" | |
| 55 "mov %0, %1\n\t" | |
| 56 | |
| 57 /* ... Toggle the ID bit in one copy and store */ | |
| 58 /* to the EFLAGS reg */ | |
| 59 "xor $0x200000, %0\n\t" | |
| 60 "push %0\n\t" | |
| 61 "popfl\n\t" | |
| 62 | |
| 63 /* ... Get the (hopefully modified) EFLAGS */ | |
| 64 "pushfl\n\t" | |
| 65 "pop %0\n\t" | |
| 66 : "=a" (a), "=c" (c) | |
| 67 : | |
| 68 : "cc" | |
| 69 ); | |
| 70 | |
| 71 if (a == c) | |
| 72 return 0; /* CPUID not supported */ | |
| 73 #endif | |
| 74 | |
| 75 cpuid(0, max_std_level, vendor.i[0], vendor.i[2], vendor.i[1]); | |
| 76 | |
| 77 if(max_std_level >= 1){ | |
| 78 cpuid(1, eax, ebx, ecx, std_caps); | |
| 79 family = ((eax>>8)&0xf) + ((eax>>20)&0xff); | |
| 80 model = ((eax>>4)&0xf) + ((eax>>12)&0xf0); | |
| 81 if (std_caps & (1<<23)) | |
| 82 rval |= FF_MM_MMX; | |
| 83 if (std_caps & (1<<25)) | |
| 84 rval |= FF_MM_MMX2 | |
| 85 #if HAVE_SSE | |
| 86 | FF_MM_SSE; | |
| 87 if (std_caps & (1<<26)) | |
| 88 rval |= FF_MM_SSE2; | |
| 89 if (ecx & 1) | |
| 90 rval |= FF_MM_SSE3; | |
| 91 if (ecx & 0x00000200 ) | |
| 92 rval |= FF_MM_SSSE3; | |
| 93 if (ecx & 0x00080000 ) | |
| 94 rval |= FF_MM_SSE4; | |
| 95 if (ecx & 0x00100000 ) | |
| 96 rval |= FF_MM_SSE42; | |
| 97 #endif | |
| 98 ; | |
| 99 } | |
| 100 | |
| 101 cpuid(0x80000000, max_ext_level, ebx, ecx, edx); | |
| 102 | |
| 103 if(max_ext_level >= 0x80000001){ | |
| 104 cpuid(0x80000001, eax, ebx, ecx, ext_caps); | |
| 105 if (ext_caps & (1<<31)) | |
| 106 rval |= FF_MM_3DNOW; | |
| 107 if (ext_caps & (1<<30)) | |
| 108 rval |= FF_MM_3DNOWEXT; | |
| 109 if (ext_caps & (1<<23)) | |
| 110 rval |= FF_MM_MMX; | |
| 111 if (ext_caps & (1<<22)) | |
| 112 rval |= FF_MM_MMX2; | |
| 113 } | |
| 114 | |
| 115 if (!strncmp(vendor.c, "GenuineIntel", 12) && | |
| 116 family == 6 && (model == 9 || model == 13 || model == 14)) { | |
| 117 /* 6/9 (pentium-m "banias"), 6/13 (pentium-m "dothan"), and 6/14 (core1
"yonah") | |
| 118 * theoretically support sse2, but it's usually slower than mmx, | |
| 119 * so let's just pretend they don't. */ | |
| 120 if (rval & FF_MM_SSE2) rval ^= FF_MM_SSE2SLOW|FF_MM_SSE2; | |
| 121 if (rval & FF_MM_SSE3) rval ^= FF_MM_SSE3SLOW|FF_MM_SSE3; | |
| 122 } | |
| 123 | |
| 124 #if 0 | |
| 125 av_log(NULL, AV_LOG_DEBUG, "%s%s%s%s%s%s%s%s%s%s%s%s\n", | |
| 126 (rval&FF_MM_MMX) ? "MMX ":"", | |
| 127 (rval&FF_MM_MMX2) ? "MMX2 ":"", | |
| 128 (rval&FF_MM_SSE) ? "SSE ":"", | |
| 129 (rval&FF_MM_SSE2) ? "SSE2 ":"", | |
| 130 (rval&FF_MM_SSE2SLOW) ? "SSE2(slow) ":"", | |
| 131 (rval&FF_MM_SSE3) ? "SSE3 ":"", | |
| 132 (rval&FF_MM_SSE3SLOW) ? "SSE3(slow) ":"", | |
| 133 (rval&FF_MM_SSSE3) ? "SSSE3 ":"", | |
| 134 (rval&FF_MM_SSE4) ? "SSE4.1 ":"", | |
| 135 (rval&FF_MM_SSE42) ? "SSE4.2 ":"", | |
| 136 (rval&FF_MM_3DNOW) ? "3DNow ":"", | |
| 137 (rval&FF_MM_3DNOWEXT) ? "3DNowExt ":""); | |
| 138 #endif | |
| 139 return rval; | |
| 140 } | |
| 141 | |
| 142 #ifdef TEST | |
| 143 int main ( void ) | |
| 144 { | |
| 145 int mm_flags; | |
| 146 mm_flags = mm_support(); | |
| 147 printf("mm_support = 0x%08X\n",mm_flags); | |
| 148 return 0; | |
| 149 } | |
| 150 #endif | |
| OLD | NEW |