/*****************************************************************************
- * osdep.h: h264 encoder
+ * osdep.h: platform-specific code
*****************************************************************************
- * Copyright (C) 2007-2008 x264 project
+ * Copyright (C) 2007-2015 x264 project
*
* Authors: Loren Merritt <lorenm@u.washington.edu>
* Laurent Aimar <fenrir@via.ecp.fr>
+ * Henrik Gramner <henrik@gramner.com>
*
* 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
* 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., 51 Franklin Street, Fifth Floor, Boston, MA 02111, USA.
+ *
+ * This program is also available under a commercial proprietary license.
+ * For more information, contact us at licensing@x264.com.
*****************************************************************************/
#ifndef X264_OSDEP_H
#define _FILE_OFFSET_BITS 64
#include <stdio.h>
#include <sys/stat.h>
+#include <inttypes.h>
+#include <stdarg.h>
#include "config.h"
-#if HAVE_STDINT_H
-#include <stdint.h>
+#ifdef __INTEL_COMPILER
+#include <mathimf.h>
#else
-#include <inttypes.h>
+#include <math.h>
#endif
#if !HAVE_LOG2F
#define log2(x) (log(x)/0.693147180559945)
#endif
-#ifdef _WIN32
-#include <io.h> // _setmode()
-#include <fcntl.h> // _O_BINARY
+#ifdef _MSC_VER
+#define inline __inline
+#define strcasecmp _stricmp
+#define strncasecmp _strnicmp
+#define strtok_r strtok_s
+#define S_ISREG(x) (((x) & S_IFMT) == S_IFREG)
+#if _MSC_VER < 1900
+#define snprintf _snprintf
+#endif
+#endif
+
+#if !defined(va_copy) && defined(__INTEL_COMPILER)
+#define va_copy(dst, src) ((dst) = (src))
#endif
-#if (SYS_OPENBSD && !defined(isfinite)) || SYS_SunOS
+#if !defined(isfinite) && (SYS_OPENBSD || SYS_SunOS)
#define isfinite finite
#endif
+
#ifdef _WIN32
-#define rename(src,dst) (unlink(dst), rename(src,dst)) // POSIX says that rename() removes the destination, but win32 doesn't.
#ifndef strtok_r
#define strtok_r(str,delim,save) strtok(str,delim)
#endif
+
+#define utf8_to_utf16( utf8, utf16 )\
+ MultiByteToWideChar( CP_UTF8, MB_ERR_INVALID_CHARS, utf8, -1, utf16, sizeof(utf16)/sizeof(wchar_t) )
+FILE *x264_fopen( const char *filename, const char *mode );
+int x264_rename( const char *oldname, const char *newname );
+#define x264_struct_stat struct _stati64
+#define x264_fstat _fstati64
+int x264_stat( const char *path, x264_struct_stat *buf );
+int x264_vfprintf( FILE *stream, const char *format, va_list arg );
+int x264_is_pipe( const char *path );
+#else
+#define x264_fopen fopen
+#define x264_rename rename
+#define x264_struct_stat struct stat
+#define x264_fstat fstat
+#define x264_stat stat
+#define x264_vfprintf vfprintf
+#define x264_is_pipe(x) 0
#endif
+#ifdef _MSC_VER
+#define DECLARE_ALIGNED( var, n ) __declspec(align(n)) var
+#else
#define DECLARE_ALIGNED( var, n ) var __attribute__((aligned(n)))
+#endif
+#define ALIGNED_32( var ) DECLARE_ALIGNED( var, 32 )
#define ALIGNED_16( var ) DECLARE_ALIGNED( var, 16 )
#define ALIGNED_8( var ) DECLARE_ALIGNED( var, 8 )
#define ALIGNED_4( var ) DECLARE_ALIGNED( var, 4 )
// - armcc can't either, but is nice enough to actually tell you so
// - Apple gcc only maintains 4 byte alignment
// - llvm can align the stack, but only in svn and (unrelated) it exposes bugs in all released GNU binutils...
+
+#define ALIGNED_ARRAY_EMU( mask, type, name, sub1, ... )\
+ uint8_t name##_u [sizeof(type sub1 __VA_ARGS__) + mask]; \
+ type (*name) __VA_ARGS__ = (void*)((intptr_t)(name##_u+mask) & ~mask)
+
#if ARCH_ARM && SYS_MACOSX
-#define ALIGNED_ARRAY_8( type, name, sub1, ... )\
- uint8_t name##_u [sizeof(type sub1 __VA_ARGS__) + 7]; \
- type (*name) __VA_ARGS__ = (void*)((intptr_t)(name##_u+7) & ~7)
+#define ALIGNED_ARRAY_8( ... ) ALIGNED_ARRAY_EMU( 7, __VA_ARGS__ )
#else
#define ALIGNED_ARRAY_8( type, name, sub1, ... )\
ALIGNED_8( type name sub1 __VA_ARGS__ )
#endif
#if ARCH_ARM
-#define ALIGNED_ARRAY_16( type, name, sub1, ... )\
- uint8_t name##_u [sizeof(type sub1 __VA_ARGS__) + 15];\
- type (*name) __VA_ARGS__ = (void*)((intptr_t)(name##_u+15) & ~15)
+#define ALIGNED_ARRAY_16( ... ) ALIGNED_ARRAY_EMU( 15, __VA_ARGS__ )
#else
#define ALIGNED_ARRAY_16( type, name, sub1, ... )\
ALIGNED_16( type name sub1 __VA_ARGS__ )
#endif
+#define EXPAND(x) x
+
+#if STACK_ALIGNMENT >= 32
+#define ALIGNED_ARRAY_32( type, name, sub1, ... )\
+ ALIGNED_32( type name sub1 __VA_ARGS__ )
+#else
+#define ALIGNED_ARRAY_32( ... ) EXPAND( ALIGNED_ARRAY_EMU( 31, __VA_ARGS__ ) )
+#endif
+
+#define ALIGNED_ARRAY_64( ... ) EXPAND( ALIGNED_ARRAY_EMU( 63, __VA_ARGS__ ) )
+
+/* For AVX2 */
+#if ARCH_X86 || ARCH_X86_64
+#define NATIVE_ALIGN 32
+#define ALIGNED_N ALIGNED_32
+#define ALIGNED_ARRAY_N ALIGNED_ARRAY_32
+#else
+#define NATIVE_ALIGN 16
+#define ALIGNED_N ALIGNED_16
+#define ALIGNED_ARRAY_N ALIGNED_ARRAY_16
+#endif
+
#if defined(__GNUC__) && (__GNUC__ > 3 || __GNUC__ == 3 && __GNUC_MINOR__ > 0)
#define UNUSED __attribute__((unused))
#define ALWAYS_INLINE __attribute__((always_inline)) inline
#define x264_constant_p(x) __builtin_constant_p(x)
#define x264_nonconstant_p(x) (!__builtin_constant_p(x))
#else
-#define UNUSED
+#ifdef _MSC_VER
+#define ALWAYS_INLINE __forceinline
+#define NOINLINE __declspec(noinline)
+#else
#define ALWAYS_INLINE inline
#define NOINLINE
+#endif
+#define UNUSED
#define MAY_ALIAS
#define x264_constant_p(x) 0
#define x264_nonconstant_p(x) 0
#endif
/* threads */
-#if SYS_BEOS
+#if HAVE_BEOSTHREAD
#include <kernel/OS.h>
#define x264_pthread_t thread_id
static inline int x264_pthread_create( x264_pthread_t *t, void *a, void *(*f)(void *), void *d )
return 0;
}
#define x264_pthread_join(t,s) { long tmp; \
- wait_for_thread(t,(s)?(long*)(*(s)):&tmp); }
-#ifndef usleep
-#define usleep(t) snooze(t)
-#endif
-#define HAVE_PTHREAD 1
+ wait_for_thread(t,(s)?(long*)(s):&tmp); }
-#elif HAVE_PTHREAD
+#elif HAVE_POSIXTHREAD
#include <pthread.h>
-#define USE_REAL_PTHREAD 1
-
-#else
-#define x264_pthread_t int
-#define x264_pthread_create(t,u,f,d) 0
-#define x264_pthread_join(t,s)
-#endif //SYS_*
-
-#if USE_REAL_PTHREAD
#define x264_pthread_t pthread_t
#define x264_pthread_create pthread_create
#define x264_pthread_join pthread_join
#define x264_pthread_attr_t pthread_attr_t
#define x264_pthread_attr_init pthread_attr_init
#define x264_pthread_attr_destroy pthread_attr_destroy
+#define x264_pthread_num_processors_np pthread_num_processors_np
#define X264_PTHREAD_MUTEX_INITIALIZER PTHREAD_MUTEX_INITIALIZER
+
+#elif HAVE_WIN32THREAD
+#include "win32thread.h"
+
#else
+#define x264_pthread_t int
+#define x264_pthread_create(t,u,f,d) 0
+#define x264_pthread_join(t,s)
+#endif //HAVE_*THREAD
+
+#if !HAVE_POSIXTHREAD && !HAVE_WIN32THREAD
#define x264_pthread_mutex_t int
#define x264_pthread_mutex_init(m,f) 0
#define x264_pthread_mutex_destroy(m)
#define X264_PTHREAD_MUTEX_INITIALIZER 0
#endif
-#define WORD_SIZE sizeof(void*)
-
-#define asm __asm__
+#if HAVE_WIN32THREAD || PTW32_STATIC_LIB
+int x264_threading_init( void );
+#else
+#define x264_threading_init() 0
+#endif
-#if !defined(_WIN64) && !defined(__LP64__)
-#if defined(__INTEL_COMPILER)
-#define BROKEN_STACK_ALIGNMENT 1 /* define it if stack is not mod16 */
+static ALWAYS_INLINE int x264_pthread_fetch_and_add( int *val, int add, x264_pthread_mutex_t *mutex )
+{
+#if HAVE_THREAD
+#if defined(__GNUC__) && (__GNUC__ > 4 || __GNUC__ == 4 && __GNUC_MINOR__ > 0) && ARCH_X86
+ return __sync_fetch_and_add( val, add );
+#else
+ x264_pthread_mutex_lock( mutex );
+ int res = *val;
+ *val += add;
+ x264_pthread_mutex_unlock( mutex );
+ return res;
#endif
+#else
+ int res = *val;
+ *val += add;
+ return res;
#endif
+}
+
+#define WORD_SIZE sizeof(void*)
+
+#define asm __asm__
#if WORDS_BIGENDIAN
#define endian_fix(x) (x)
#define endian_fix32(x) (x)
#define endian_fix16(x) (x)
#else
-#if defined(__GNUC__) && HAVE_MMX
+#if HAVE_X86_INLINE_ASM && HAVE_MMX
static ALWAYS_INLINE uint32_t endian_fix32( uint32_t x )
{
asm("bswap %0":"+r"(x));
return (x<<24) + ((x<<8)&0xff0000) + ((x>>8)&0xff00) + (x>>24);
}
#endif
-#if defined(__GNUC__) && ARCH_X86_64
+#if HAVE_X86_INLINE_ASM && ARCH_X86_64
static ALWAYS_INLINE uint64_t endian_fix64( uint64_t x )
{
asm("bswap %0":"+r"(x));
}
#endif
+/* For values with 4 bits or less. */
+static int ALWAYS_INLINE x264_ctz_4bit( uint32_t x )
+{
+ static uint8_t lut[16] = {4,0,1,0,2,0,1,0,3,0,1,0,2,0,1,0};
+ return lut[x];
+}
+
#if defined(__GNUC__) && (__GNUC__ > 3 || __GNUC__ == 3 && __GNUC_MINOR__ > 3)
#define x264_clz(x) __builtin_clz(x)
#define x264_ctz(x) __builtin_ctz(x)
}
#endif
-#if defined(__GNUC__) && HAVE_MMX
+#if HAVE_X86_INLINE_ASM && HAVE_MMX
/* Don't use __builtin_prefetch; even as recent as 4.3.4, GCC seems incapable of
* using complex address modes properly unless we use inline asm. */
static ALWAYS_INLINE void x264_prefetch( void *p )
#define x264_prefetch(x)
#endif
-#if USE_REAL_PTHREAD
-#if SYS_MINGW
+#if HAVE_POSIXTHREAD
+#if SYS_WINDOWS
#define x264_lower_thread_priority(p)\
{\
x264_pthread_t handle = pthread_self();\
#else
#include <unistd.h>
#define x264_lower_thread_priority(p) { UNUSED int nice_ret = nice(p); }
-#endif /* USE_REAL_PTHREAD */
+#endif /* SYS_WINDOWS */
+#elif HAVE_WIN32THREAD
+#define x264_lower_thread_priority(p) SetThreadPriority( GetCurrentThread(), X264_MAX( -2, -p ) )
#else
#define x264_lower_thread_priority(p)
#endif
-static inline uint8_t x264_is_regular_file( FILE *filehandle )
+static inline int x264_is_regular_file( FILE *filehandle )
+{
+ x264_struct_stat file_stat;
+ if( x264_fstat( fileno( filehandle ), &file_stat ) )
+ return 1;
+ return S_ISREG( file_stat.st_mode );
+}
+
+static inline int x264_is_regular_file_path( const char *filename )
{
- struct stat file_stat;
- if( fstat( fileno( filehandle ), &file_stat ) )
- return 0;
+ x264_struct_stat file_stat;
+ if( x264_stat( filename, &file_stat ) )
+ return !x264_is_pipe( filename );
return S_ISREG( file_stat.st_mode );
}