Function generic_reduce

include/boost/compute/algorithm/reduce.hpp:161–203

// Space complexity: O( ceil(n / 2 / 256) )
template<class InputIterator, class OutputIterator, class BinaryFunction>
inline void generic_reduce(InputIterator first,
                           InputIterator last,
                           OutputIterator result,
                           BinaryFunction function,
                           command_queue &queue)
{
    typedef typename
        std::iterator_traits<InputIterator>::value_type
        input_type;
    typedef typename
        boost::compute::result_of<BinaryFunction(input_type, input_type)>::type
        result_type;

    const device &device = queue.get_device();
    const context &context = queue.get_context();

    size_t count = detail::iterator_range_size(first, last);

    if(device.type() & device::cpu){
        array<result_type, 1> value(context);
        detail::reduce_on_cpu(first, last, value.begin(), function, queue);
        boost::compute::copy_n(value.begin(), 1, result, queue);
    }
    else {
        size_t block_size = 256;

        // first pass
        vector<result_type> results = detail::block_reduce(first,
                                                           count,
                                                           block_size,
                                                           function,
                                                           queue);

        if(results.size() > 1){
            detail::inplace_reduce(results.begin(),
                                   results.end(),
                                   function,
                                   queue);
        }

        boost::compute::copy_n(results.begin(), 1, result, queue);
    }
}

template<class InputIterator, class OutputIterator, class T>
inline void dispatch_reduce(InputIterator first,