panthema / 2013 / parallel-string-sorting / parallel-string-sorting-0.5 / minitbb / tbb / machine / linux_ia32.h (Download File)
/*
    Copyright 2005-2012 Intel Corporation.  All Rights Reserved.

    This file is part of Threading Building Blocks.

    Threading Building Blocks is free software; you can redistribute it
    and/or modify it under the terms of the GNU General Public License
    version 2 as published by the Free Software Foundation.

    Threading Building Blocks 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 Threading Building Blocks; if not, write to the Free Software
    Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA

    As a special exception, you may use this file as part of a free software
    library without restriction.  Specifically, if other files instantiate
    templates or use macros or inline functions from this file, or you compile
    this file and link it with other files to produce an executable, this
    file does not by itself cause the resulting executable to be covered by
    the GNU General Public License.  This exception does not however
    invalidate any other reasons why the executable file might be covered by
    the GNU General Public License.
*/

#if !defined(__TBB_machine_H) || defined(__TBB_machine_linux_ia32_H)
#error Do not #include this internal file directly; use public TBB headers instead.
#endif

#define __TBB_machine_linux_ia32_H

#include <stdint.h>
#include "gcc_ia32_common.h"

#define __TBB_WORDSIZE 4
#define __TBB_BIG_ENDIAN 0

#define __TBB_compiler_fence() __asm__ __volatile__("": : :"memory")
#define __TBB_control_consistency_helper() __TBB_compiler_fence()
#define __TBB_acquire_consistency_helper() __TBB_compiler_fence()
#define __TBB_release_consistency_helper() __TBB_compiler_fence()
#define __TBB_full_memory_fence()          __asm__ __volatile__("mfence": : :"memory")

#if __TBB_ICC_ASM_VOLATILE_BROKEN
#define __TBB_VOLATILE
#else
#define __TBB_VOLATILE volatile
#endif

#define __TBB_MACHINE_DEFINE_ATOMICS(S,T,X,R)                                        \
static inline T __TBB_machine_cmpswp##S (volatile void *ptr, T value, T comparand )  \
{                                                                                    \
    T result;                                                                        \
                                                                                     \
    __asm__ __volatile__("lock\ncmpxchg" X " %2,%1"                                  \
                          : "=a"(result), "=m"(*(__TBB_VOLATILE T*)ptr)              \
                          : "q"(value), "0"(comparand), "m"(*(__TBB_VOLATILE T*)ptr) \
                          : "memory");                                               \
    return result;                                                                   \
}                                                                                    \
                                                                                     \
static inline T __TBB_machine_fetchadd##S(volatile void *ptr, T addend)              \
{                                                                                    \
    T result;                                                                        \
    __asm__ __volatile__("lock\nxadd" X " %0,%1"                                     \
                          : R (result), "=m"(*(__TBB_VOLATILE T*)ptr)                \
                          : "0"(addend), "m"(*(__TBB_VOLATILE T*)ptr)                \
                          : "memory");                                               \
    return result;                                                                   \
}                                                                                    \
                                                                                     \
static inline  T __TBB_machine_fetchstore##S(volatile void *ptr, T value)            \
{                                                                                    \
    T result;                                                                        \
    __asm__ __volatile__("lock\nxchg" X " %0,%1"                                     \
                          : R (result), "=m"(*(__TBB_VOLATILE T*)ptr)                \
                          : "0"(value), "m"(*(__TBB_VOLATILE T*)ptr)                 \
                          : "memory");                                               \
    return result;                                                                   \
}                                                                                    \
                                                                                     
__TBB_MACHINE_DEFINE_ATOMICS(1,int8_t,"","=q")
__TBB_MACHINE_DEFINE_ATOMICS(2,int16_t,"","=r")
__TBB_MACHINE_DEFINE_ATOMICS(4,int32_t,"l","=r")

#if __INTEL_COMPILER
#pragma warning( push )
// reference to EBX in a function requiring stack alignment
#pragma warning( disable: 998 )
#endif

static inline int64_t __TBB_machine_cmpswp8 (volatile void *ptr, int64_t value, int64_t comparand ) {
#if __TBB_GCC_BUILTIN_ATOMICS_PRESENT
    return __sync_val_compare_and_swap( reinterpret_cast<volatile int64_t*>(ptr), comparand, value );
#else /* !__TBB_GCC_BUILTIN_ATOMICS_PRESENT */
    //TODO: look like ICC 13.0 has some issues with this code, investigate it more deeply
    int64_t result;
    union {
        int64_t i64;
        int32_t i32[2];
    };
    i64 = value;
#if __PIC__ 
    /* compiling position-independent code */
    // EBX register preserved for compliance with position-independent code rules on IA32
    int32_t tmp;
    __asm__ __volatile__ (
            "movl  %%ebx,%2\n\t"
            "movl  %5,%%ebx\n\t"
#if __GNUC__==3
            "lock\n\t cmpxchg8b %1\n\t"
#else
            "lock\n\t cmpxchg8b (%3)\n\t"
#endif
            "movl  %2,%%ebx"
             : "=A"(result)
             , "=m"(*(__TBB_VOLATILE int64_t *)ptr)
             , "=m"(tmp)
#if __GNUC__==3
             : "m"(*(__TBB_VOLATILE int64_t *)ptr)
#else
             : "SD"(ptr)
#endif
             , "0"(comparand)
             , "m"(i32[0]), "c"(i32[1])
             : "memory"
#if __INTEL_COMPILER
             ,"ebx"
#endif
    );
#else /* !__PIC__ */
    __asm__ __volatile__ (
            "lock\n\t cmpxchg8b %1\n\t"
             : "=A"(result), "=m"(*(__TBB_VOLATILE int64_t *)ptr)
             : "m"(*(__TBB_VOLATILE int64_t *)ptr)
             , "0"(comparand)
             , "b"(i32[0]), "c"(i32[1])
             : "memory"
    );
#endif /* __PIC__ */
    return result;
#endif /* !__TBB_GCC_BUILTIN_ATOMICS_PRESENT */
}

#if __INTEL_COMPILER
#pragma warning( pop )
#endif // warning 998 is back

static inline void __TBB_machine_or( volatile void *ptr, uint32_t addend ) {
    __asm__ __volatile__("lock\norl %1,%0" : "=m"(*(__TBB_VOLATILE uint32_t *)ptr) : "r"(addend), "m"(*(__TBB_VOLATILE uint32_t *)ptr) : "memory");
}

static inline void __TBB_machine_and( volatile void *ptr, uint32_t addend ) {
    __asm__ __volatile__("lock\nandl %1,%0" : "=m"(*(__TBB_VOLATILE uint32_t *)ptr) : "r"(addend), "m"(*(__TBB_VOLATILE uint32_t *)ptr) : "memory");
}

//TODO: Check if it possible and profitable for IA-32 on (Linux and Windows)
//to use of 64-bit load/store via floating point registers together with full fence
//for sequentially consistent load/store, instead of CAS.

#if __clang__
#define __TBB_fildq  "fildll"
#define __TBB_fistpq "fistpll"
#else
#define __TBB_fildq  "fildq"
#define __TBB_fistpq "fistpq"
#endif

static inline int64_t __TBB_machine_load8 (const volatile void *ptr) {
    int64_t result;
    if( ((uint32_t)ptr&7u)==0 ) {
        // Aligned load
        __asm__ __volatile__ ( __TBB_fildq  " %1\n\t"
                               __TBB_fistpq " %0" :  "=m"(result) : "m"(*(const __TBB_VOLATILE uint64_t*)ptr) : "memory" );
    } else {
        // Unaligned load
        result = __TBB_machine_cmpswp8(const_cast<void*>(ptr),0,0);
    }
    return result;
}

//! Handles misaligned 8-byte store
/** Defined in tbb_misc.cpp */
extern "C" void __TBB_machine_store8_slow( volatile void *ptr, int64_t value );
extern "C" void __TBB_machine_store8_slow_perf_warning( volatile void *ptr );

static inline void __TBB_machine_store8(volatile void *ptr, int64_t value) {
    if( ((uint32_t)ptr&7u)==0 ) {
        // Aligned store
        __asm__ __volatile__ ( __TBB_fildq  " %1\n\t"
                               __TBB_fistpq " %0" :  "=m"(*(__TBB_VOLATILE int64_t*)ptr) : "m"(value) : "memory" );
    } else {
        // Unaligned store
#if TBB_USE_PERFORMANCE_WARNINGS
        __TBB_machine_store8_slow_perf_warning(ptr);
#endif /* TBB_USE_PERFORMANCE_WARNINGS */
        __TBB_machine_store8_slow(ptr,value);
    }
}
 
// Machine specific atomic operations
#define __TBB_AtomicOR(P,V) __TBB_machine_or(P,V)
#define __TBB_AtomicAND(P,V) __TBB_machine_and(P,V)

#define __TBB_USE_GENERIC_DWORD_FETCH_ADD                   1
#define __TBB_USE_GENERIC_DWORD_FETCH_STORE                 1
#define __TBB_USE_FETCHSTORE_AS_FULL_FENCED_STORE           1
#define __TBB_USE_GENERIC_HALF_FENCED_LOAD_STORE            1
#define __TBB_USE_GENERIC_RELAXED_LOAD_STORE                1
#define __TBB_USE_GENERIC_SEQUENTIAL_CONSISTENCY_LOAD_STORE 1