Parallel "reduce"

reduce.cpp

template <class RandomAccessIterator, class T, class Func>
T Reduce(RandomAccessIterator first, RandomAccessIterator last, const T& initial_value, Func func) {
    std::vector<std::thread> workers;
    std::atomic<T> cur_value(initial_value);
    auto dist = std::distance(first, last);
    auto seglen = dist / 7;
    auto rem = dist % 7;
    if (seglen != 0) {

        for (size_t i = 0; i < 7; i++) {
            workers.emplace_back([&first, &last, &i, &func, &cur_value, &seglen] {
                auto it = first + i * seglen;
                while (it != std::min(last, first + (i + 1) * seglen)) {
                    cur_value = func(cur_value, *it++);
                }
            });
        }
    }
    workers.emplace_back([&first, &last, &rem, &func, &cur_value, &seglen] {
        auto it = last - rem;
        while (it != last) {
            cur_value = func(cur_value, *it++);
        }
    });

    for (auto& worker : workers) {
        worker.join();
    }

    return cur_value;
}