利用union实现内存池（boost::quick_allocator）

// 根据size的不同， 可以实现不同大小的多个内存池

template<unsigned size, unsigned align_> union freeblock

{

typedef typename boost::type_with_alignment<align_>::type aligner_type;

aligner_type aligner;

// 这个地方利用union真的很棒

char bytes[size];

freeblock * next;

};

template<unsigned size, unsigned align_> struct allocator_impl

{

typedef freeblock<size, align_> block;

// It may seem odd to use such small pages.

//

// However, on a typical Windows implementation that uses

// the OS allocator, "normal size" pages interact with the

// "ordinary" operator new, slowing it down dramatically.

//

// 512 byte pages are handled by the small object allocator,

// and don't interfere with ::new.

//

// The other alternative is to use much bigger pages (1M.)

//

// It is surprisingly easy to hit pathological behavior by

// varying the page size. g++ 2.96 on Red Hat Linux 7.2,

// for example, passionately dislikes 496. 512 seems OK.

#if defined(BOOST_QA_PAGE_SIZE)

enum { items_per_page = BOOST_QA_PAGE_SIZE / size };

#else

enum { items_per_page = 512 / size }; // 1048560 / size

#endif

#ifdef BOOST_HAS_THREADS

static lightweight_mutex & mutex()

{

static freeblock< sizeof( lightweight_mutex ), boost::alignment_of< lightweight_mutex >::value > fbm;

static lightweight_mutex * pm = new( &fbm ) lightweight_mutex;

return *pm;

}

static lightweight_mutex * mutex_init;

#endif

static block * free;

static block * page;

static unsigned last;

static inline void * alloc()

{

#ifdef BOOST_HAS_THREADS

lightweight_mutex::scoped_lock lock( mutex() );

#endif

if(block * x = free)

{

free = x->next;

return x;

}

else

{

if(last == items_per_page)

{

// "Listen to me carefully: there is no memory leak"

// -- Scott Meyers, Eff C++ 2nd Ed Item 10

page = ::new block[items_per_page];

last = 0;

}

return &page[last++];

}

}

static inline void * alloc(std::size_t n)

{

if(n != size) // class-specific new called for a derived object

{

return ::operator new(n);

}

else

{

#ifdef BOOST_HAS_THREADS

lightweight_mutex::scoped_lock lock( mutex() );

#endif

if(block * x = free)

{

free = x->next;

return x;

}

else

{

if(last == items_per_page)

{

page = ::new block[items_per_page];

last = 0;

}

return &page[last++];

}

}

}

static inline void dealloc(void * pv)

{

if(pv != 0) // 18.4.1.1/13

{

#ifdef BOOST_HAS_THREADS

lightweight_mutex::scoped_lock lock( mutex() );

#endif

block * pb = static_cast<block *>(pv);

pb->next = free;

free = pb;

}

}

static inline void dealloc(void * pv, std::size_t n)

{

if(n != size) // class-specific delete called for a derived object

{

::operator delete(pv);

}

else if(pv != 0) // 18.4.1.1/13

{

#ifdef BOOST_HAS_THREADS

lightweight_mutex::scoped_lock lock( mutex() );

#endif

block * pb = static_cast<block *>(pv);

pb->next = free;

free = pb;

}

}

};