X-Git-Url: https://git.sesse.net/?a=blobdiff_plain;f=src%2Fmisc%2Fcpu.c;h=124876d7adfc02cebbb68f343ce2a58baa5fd9cd;hb=79559bdf7ff37fb54364a3ef7917b1e5481dc4c2;hp=c73ba7f42c3db0f888853c16830bfefb24fe3791;hpb=bc061e8455fa11eec3167681d53f9df3974b13a5;p=vlc diff --git a/src/misc/cpu.c b/src/misc/cpu.c index c73ba7f42c..124876d7ad 100644 --- a/src/misc/cpu.c +++ b/src/misc/cpu.c @@ -1,8 +1,8 @@ /***************************************************************************** * cpu.c: CPU detection code ***************************************************************************** - * Copyright (C) 1998-2002 VideoLAN - * $Id: cpu.c,v 1.9 2002/11/11 14:39:12 sam Exp $ + * Copyright (C) 1998-2004 the VideoLAN team + * $Id$ * * Authors: Samuel Hocevar * Christophe Massiot @@ -20,27 +20,28 @@ * * You should have received a copy of the GNU General Public License * along with this program; if not, write to the Free Software - * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111, USA. + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston MA 02110-1301, USA. *****************************************************************************/ /***************************************************************************** * Preamble *****************************************************************************/ -#include +#ifdef HAVE_CONFIG_H +# include "config.h" +#endif + +#include #ifdef HAVE_SIGNAL_H # include /* SIGHUP, SIGINT, SIGKILL */ # include /* longjmp, setjmp */ #endif -#ifdef SYS_DARWIN -# include /* AltiVec detection */ -# include /* some day the header files||compiler * - will define it for us */ -# include -#endif +#include "libvlc.h" -#include "vlc_cpu.h" +#if defined(__APPLE__) && (defined(__ppc__) || defined(__ppc64__)) +#include +#endif /***************************************************************************** * Local prototypes @@ -55,8 +56,8 @@ static void SigHandler ( int ); #ifdef HAVE_SIGNAL_H static jmp_buf env; static int i_illegal; -#if defined( __i386__ ) -static char *psz_capability; +#if defined( __i386__ ) || defined( __x86_64__ ) +static const char *psz_capability; #endif #endif @@ -69,54 +70,47 @@ uint32_t CPUCapabilities( void ) { volatile uint32_t i_capabilities = CPU_CAPABILITY_NONE; -#if defined( SYS_DARWIN ) - struct host_basic_info hi; - kern_return_t ret; - host_name_port_t host; - - int i_size; - char *psz_name, *psz_subname; +#if defined(__APPLE__) && (defined(__ppc__) || defined(__ppc64__)) + int selectors[2] = { CTL_HW, HW_VECTORUNIT }; + int i_has_altivec = 0; + size_t i_length = sizeof( i_has_altivec ); + int i_error = sysctl( selectors, 2, &i_has_altivec, &i_length, NULL, 0); i_capabilities |= CPU_CAPABILITY_FPU; - /* Should 'never' fail? */ - host = mach_host_self(); - - i_size = sizeof( hi ) / sizeof( int ); - ret = host_info( host, HOST_BASIC_INFO, ( host_info_t )&hi, &i_size ); - - if( ret != KERN_SUCCESS ) - { - fprintf( stderr, "error: couldn't get CPU information\n" ); - return i_capabilities; - } - - slot_name( hi.cpu_type, hi.cpu_subtype, &psz_name, &psz_subname ); - /* FIXME: need better way to detect newer proccessors. - * could do strncmp(a,b,5), but that's real ugly */ - if( !strcmp(psz_name, "ppc7400") || !strcmp(psz_name, "ppc7450") ) - { + if( i_error == 0 && i_has_altivec != 0 ) i_capabilities |= CPU_CAPABILITY_ALTIVEC; - } return i_capabilities; -#elif defined( __i386__ ) +#elif defined( __i386__ ) || defined( __x86_64__ ) volatile unsigned int i_eax, i_ebx, i_ecx, i_edx; - volatile vlc_bool_t b_amd; + volatile bool b_amd; /* Needed for x86 CPU capabilities detection */ -# define cpuid( a ) \ - asm volatile ( "pushl %%ebx\n\t" \ - "cpuid\n\t" \ - "movl %%ebx,%1\n\t" \ - "popl %%ebx\n\t" \ - : "=a" ( i_eax ), \ - "=r" ( i_ebx ), \ - "=c" ( i_ecx ), \ - "=d" ( i_edx ) \ - : "a" ( a ) \ - : "cc" ); +# if defined( __x86_64__ ) +# define cpuid( reg ) \ + asm volatile ( "cpuid\n\t" \ + "movl %%ebx,%1\n\t" \ + : "=a" ( i_eax ), \ + "=b" ( i_ebx ), \ + "=c" ( i_ecx ), \ + "=d" ( i_edx ) \ + : "a" ( reg ) \ + : "cc" ); +# else +# define cpuid( reg ) \ + asm volatile ( "push %%ebx\n\t" \ + "cpuid\n\t" \ + "movl %%ebx,%1\n\t" \ + "pop %%ebx\n\t" \ + : "=a" ( i_eax ), \ + "=r" ( i_ebx ), \ + "=c" ( i_ecx ), \ + "=d" ( i_edx ) \ + : "a" ( reg ) \ + : "cc" ); +# endif # if defined( CAN_COMPILE_SSE ) || defined ( CAN_COMPILE_3DNOW ) \ && defined( HAVE_SIGNAL_H ) @@ -125,18 +119,19 @@ uint32_t CPUCapabilities( void ) i_capabilities |= CPU_CAPABILITY_FPU; - /* test for a 486 CPU */ - asm volatile ( "pushl %%ebx\n\t" - "pushfl\n\t" - "popl %%eax\n\t" +# if defined( __i386__ ) + /* check if cpuid instruction is supported */ + asm volatile ( "push %%ebx\n\t" + "pushf\n\t" + "pop %%eax\n\t" "movl %%eax, %%ebx\n\t" "xorl $0x200000, %%eax\n\t" - "pushl %%eax\n\t" - "popfl\n\t" - "pushfl\n\t" - "popl %%eax\n\t" + "push %%eax\n\t" + "popf\n\t" + "pushf\n\t" + "pop %%eax\n\t" "movl %%ebx,%1\n\t" - "popl %%ebx\n\t" + "pop %%ebx\n\t" : "=a" ( i_eax ), "=r" ( i_ebx ) : @@ -144,12 +139,15 @@ uint32_t CPUCapabilities( void ) if( i_eax == i_ebx ) { -# if defined( CAN_COMPILE_SSE ) || defined ( CAN_COMPILE_3DNOW ) \ - && defined( HAVE_SIGNAL_H ) +# if defined( CAN_COMPILE_SSE ) || defined ( CAN_COMPILE_3DNOW ) \ + && defined( HAVE_SIGNAL_H ) signal( SIGILL, pf_sigill ); -# endif +# endif return i_capabilities; } +# else + /* x86_64 supports cpuid instruction, so we dont need to check it */ +# endif i_capabilities |= CPU_CAPABILITY_486; @@ -208,6 +206,26 @@ uint32_t CPUCapabilities( void ) # endif } + if( i_edx & 0x04000000 ) + { +# if defined(CAN_COMPILE_SSE) + /* We test if OS supports the SSE instructions */ + psz_capability = "SSE2"; + i_illegal = 0; + + if( setjmp( env ) == 0 ) + { + /* Test a SSE2 instruction */ + __asm__ __volatile__ ( "movupd %%xmm0, %%xmm0\n" : : ); + } + + if( i_illegal == 0 ) + { + i_capabilities |= CPU_CAPABILITY_SSE2; + } +# endif + } + /* test for additional capabilities */ cpuid( 0x80000000 ); @@ -253,7 +271,7 @@ uint32_t CPUCapabilities( void ) # endif return i_capabilities; -#elif defined( __powerpc__ ) +#elif defined( __powerpc__ ) || defined( __ppc__ ) || defined( __ppc64__ ) # ifdef CAN_COMPILE_ALTIVEC && defined( HAVE_SIGNAL_H ) void (*pf_sigill) (int) = signal( SIGILL, SigHandler ); @@ -285,6 +303,10 @@ uint32_t CPUCapabilities( void ) i_capabilities |= CPU_CAPABILITY_FPU; return i_capabilities; +#elif defined( _MSC_VER ) && !defined( UNDER_CE ) + i_capabilities |= CPU_CAPABILITY_FPU; + return i_capabilities; + #else /* default behaviour */ return i_capabilities; @@ -313,7 +335,7 @@ static void SigHandler( int i_signal ) "operating system.\n", psz_capability ); fprintf( stderr, " some optimizations will be disabled unless " "you upgrade your OS\n" ); -# if defined( SYS_LINUX ) +# if defined( __linux__ ) fprintf( stderr, " (for instance Linux kernel 2.4.x or later)\n" ); # endif #endif @@ -322,3 +344,41 @@ static void SigHandler( int i_signal ) } #endif + +uint32_t cpu_flags = 0; + + +/***************************************************************************** + * vlc_CPU: get pre-computed CPU capability flags + ****************************************************************************/ +unsigned vlc_CPU (void) +{ + return cpu_flags; +} + +static vlc_memcpy_t pf_vlc_memcpy = memcpy; +static vlc_memset_t pf_vlc_memset = memset; + +void vlc_fastmem_register (vlc_memcpy_t cpy, vlc_memset_t set) +{ + if (cpy) + pf_vlc_memcpy = cpy; + if (set) + pf_vlc_memset = set; +} + +/** + * vlc_memcpy: fast CPU-dependent memcpy + */ +void *vlc_memcpy (void *tgt, const void *src, size_t n) +{ + return pf_vlc_memcpy (tgt, src, n); +} + +/** + * vlc_memset: fast CPU-dependent memset + */ +void *vlc_memset (void *tgt, int c, size_t n) +{ + return pf_vlc_memset (tgt, c, n); +}