C++并发编程——在运行时选择线程数量

在编写多线程程序时，运行多少线程比较合适呢？线程并不是越多越好，理论上，硬件支持多少线程数，就开多少个线程比较合适，有的比如完成端口IOCP中建议开2倍线程数，因为考虑到有些线程可能会挂起等情况。但最重要的一条，首先要获取当前硬件支持的线程数，通常情况下为CPU核数。

std::thread::hardware_concurrency(); //获取当前CPU核心数量

代码示例：
以下代码为std::accumulate的简单并行版本实现，通过将大量的累加操作，分配给多个线程去计算，最后将各个线程计算的结果累加，得出最终结果。真正的并行计算任务分割是很麻烦的，这里并不需要考虑线程的同步等问题。

template<typename Iterator,typename T>struct accumulate_block{    void operator()(Iterator first,Iterator last,T& result)    {        result = std::accumulate(first,last,result);    }};template<typename Iterator,typename T>T parallel_accumulate(Iterator first,Iterator last,T init){    unsigned long const length=std::distance(first,last);    if(!length)        return init;    unsigned long const min_per_thread=25;    unsigned long const max_threads=(length+min_per_thread-1)/min_per_thread;   //获取最大线程数量    unsigned long const hardware_threads=std::thread::hardware_concurrency();   //获取当前CPU核心数量    unsigned long const num_threads=std::min(hardware_threads!=0?hardware_threads:2,max_threads);//运行线程数量    unsigned long const block_size=length/num_threads;    std::vector<T> results(num_threads);    Iterator block_start=first;    std::vector<std::thread> v_threads(num_threads-1);    for(unsigned long i=0;i<num_threads-1;++i)    {        Iterator block_end=block_start;        std::advance(block_end,block_size);        v_threads[i]=std::thread(accumulate_block<Iterator,T>(),block_start,block_end,std::ref(results[i]));        block_start=block_end;    }    accumulate_block<Iterator,T>()(block_start,last,results[num_threads-1]);    //计算剩下的数,相当于在主线程中计算    std::for_each(v_threads.begin(),v_threads.end(),std::mem_fn(&std::thread::join));//等待所有线程计算完成    return std::accumulate(results.begin(),results.end(),init);}int _tmain(int argc, _TCHAR* argv[]){    std::vector<int> v(100000);    std::iota(v.begin(),v.end(),1);    long long sum=parallel_accumulate(v.begin(),v.end(),0);    cout<<"sum="<<sum<<endl;    return 0;}

相关STL源码：

//std::distance源码template<class _BidIt,    class _Diff> inline        void _Distance2(_BidIt _First, _BidIt _Last, _Diff& _Off,            bidirectional_iterator_tag)    {   // add to _Off distance between bidirectional iterators (redundant)    for (; _First != _Last; ++_First)        ++_Off;    }template<class _InIt> inline    typename iterator_traits<_InIt>::difference_type        distance(_InIt _First, _InIt _Last)    {   // return distance between iterators    typename iterator_traits<_InIt>::difference_type _Off = 0;    _Distance2(_First, _Last, _Off, _Iter_cat(_First));    return (_Off);    }

//std::advance源代码    // TEMPLATE FUNCTION advancetemplate<class _InIt,    class _Diff> inline    void _Advance(_InIt& _Where, _Diff _Off, input_iterator_tag)    {   // increment iterator by offset, input iterators #if _ITERATOR_DEBUG_LEVEL == 2    if (_Off < 0)        _DEBUG_ERROR("negative offset in advance"); #endif /* _ITERATOR_DEBUG_LEVEL == 2 */    for (; 0 < _Off; --_Off)        ++_Where;    }template<class _FwdIt,    class _Diff> inline    void _Advance(_FwdIt& _Where, _Diff _Off, forward_iterator_tag)    {   // increment iterator by offset, forward iterators #if _ITERATOR_DEBUG_LEVEL == 2    if (_Off < 0)        _DEBUG_ERROR("negative offset in advance"); #endif /* _ITERATOR_DEBUG_LEVEL == 2 */    for (; 0 < _Off; --_Off)        ++_Where;    }template<class _BidIt,    class _Diff> inline    void _Advance(_BidIt& _Where, _Diff _Off, bidirectional_iterator_tag)    {   // increment iterator by offset, bidirectional iterators    for (; 0 < _Off; --_Off)        ++_Where;    for (; _Off < 0; ++_Off)        --_Where;    }template<class _RanIt,    class _Diff> inline    void _Advance(_RanIt& _Where, _Diff _Off, random_access_iterator_tag)    {   // increment iterator by offset, random-access iterators    _Where += _Off;    }template<class _InIt,    class _Diff> inline    void advance(_InIt& _Where, _Diff _Off)    {   // increment iterator by offset, arbitrary iterators    _Advance(_Where, _Off, _Iter_cat(_Where));    }

//获取硬件线程数量    static unsigned int hardware_concurrency() _NOEXCEPT        {   // return number of hardware thread contexts        return (::Concurrency::details::_GetConcurrency());        }