/*
    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_windows_ia32_H)
#error Do not #include this internal file directly; use public TBB headers instead.
#endif

#define __TBB_machine_windows_ia32_H

#include "msvc_ia32_common.h"

#define __TBB_WORDSIZE 4
#define __TBB_BIG_ENDIAN 0

#if __INTEL_COMPILER && (__INTEL_COMPILER < 1100)
    #define __TBB_compiler_fence()    __asm { __asm nop }
    #define __TBB_full_memory_fence() __asm { __asm mfence }
#elif _MSC_VER >= 1300 || __INTEL_COMPILER
    #pragma intrinsic(_ReadWriteBarrier)
    #pragma intrinsic(_mm_mfence)
    #define __TBB_compiler_fence()    _ReadWriteBarrier()
    #define __TBB_full_memory_fence() _mm_mfence()
#else
    #error Unsupported compiler - need to define __TBB_{control,acquire,release}_consistency_helper to support it
#endif

#define __TBB_control_consistency_helper() __TBB_compiler_fence()
#define __TBB_acquire_consistency_helper() __TBB_compiler_fence()
#define __TBB_release_consistency_helper() __TBB_compiler_fence()

#if defined(_MSC_VER) && !defined(__INTEL_COMPILER)
    // Workaround for overzealous compiler warnings in /Wp64 mode
    #pragma warning (push)
    #pragma warning (disable: 4244 4267)
#endif

extern "C" {
    __int64 __TBB_EXPORTED_FUNC __TBB_machine_cmpswp8 (volatile void *ptr, __int64 value, __int64 comparand );
    __int64 __TBB_EXPORTED_FUNC __TBB_machine_fetchadd8 (volatile void *ptr, __int64 addend );
    __int64 __TBB_EXPORTED_FUNC __TBB_machine_fetchstore8 (volatile void *ptr, __int64 value );
    void __TBB_EXPORTED_FUNC __TBB_machine_store8 (volatile void *ptr, __int64 value );
    __int64 __TBB_EXPORTED_FUNC __TBB_machine_load8 (const volatile void *ptr);
}

//TODO: use _InterlockedXXX intrinsics as they available since VC 2005
#define __TBB_MACHINE_DEFINE_ATOMICS(S,T,U,A,C) \
static inline T __TBB_machine_cmpswp##S ( volatile void * ptr, U value, U comparand ) { \
    T result; \
    volatile T *p = (T *)ptr; \
    __asm \
    { \
       __asm mov edx, p \
       __asm mov C , value \
       __asm mov A , comparand \
       __asm lock cmpxchg [edx], C \
       __asm mov result, A \
    } \
    return result; \
} \
\
static inline T __TBB_machine_fetchadd##S ( volatile void * ptr, U addend ) { \
    T result; \
    volatile T *p = (T *)ptr; \
    __asm \
    { \
        __asm mov edx, p \
        __asm mov A, addend \
        __asm lock xadd [edx], A \
        __asm mov result, A \
    } \
    return result; \
}\
\
static inline T __TBB_machine_fetchstore##S ( volatile void * ptr, U value ) { \
    T result; \
    volatile T *p = (T *)ptr; \
    __asm \
    { \
        __asm mov edx, p \
        __asm mov A, value \
        __asm lock xchg [edx], A \
        __asm mov result, A \
    } \
    return result; \
}


__TBB_MACHINE_DEFINE_ATOMICS(1, __int8, __int8, al, cl)
__TBB_MACHINE_DEFINE_ATOMICS(2, __int16, __int16, ax, cx)
__TBB_MACHINE_DEFINE_ATOMICS(4, ptrdiff_t, ptrdiff_t, eax, ecx)

#undef __TBB_MACHINE_DEFINE_ATOMICS

static inline void __TBB_machine_OR( volatile void *operand, __int32 addend ) {
   __asm 
   {
       mov eax, addend
       mov edx, [operand]
       lock or [edx], eax
   }
}

static inline void __TBB_machine_AND( volatile void *operand, __int32 addend ) {
   __asm 
   {
       mov eax, addend
       mov edx, [operand]
       lock and [edx], eax
   }
}

#define __TBB_AtomicOR(P,V) __TBB_machine_OR(P,V)
#define __TBB_AtomicAND(P,V) __TBB_machine_AND(P,V)

//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.
#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


#if defined(_MSC_VER) && !defined(__INTEL_COMPILER)
    #pragma warning (pop)
#endif // warnings 4244, 4267 are back