std::async, C++ (vc++)

Language: Layout:

#include <iostream>
#include <string>
#include <future>
#include <mutex>
#include <algorithm>
#include <vector>
#include <numeric>
#include <atomic>

template < typename RA_ITERATOR, typename FN >
void parallel_for( RA_ITERATOR begin, RA_ITERATOR end, FN fn, std::size_t seg_size = 1024*128 )
{
    const std::size_t n = end - begin ;

#ifndef NDEBUG
    {
       static std::mutex lock ;
       std::lock_guard<std::mutex> guard(lock) ;
       std::cout << "parallel_for: n == " << n << " thread: " << std::this_thread::get_id() << '\n' ;
    }
    #endif // NDEBUG

if( n < seg_size ) std::for_each( begin, end, fn ) ;

else
    {
        auto future = std::async( std::launch::async, parallel_for<RA_ITERATOR,FN>, begin, begin+n/2, fn, seg_size ) ;
        parallel_for( begin+n/2, end, fn, seg_size ) ;
        future.wait() ;
    }
}

int main ()
{
    const long long N = 1'000'000 ;

std::vector<int> vec(N) ;
    std::iota( vec.begin(), vec.end(), 1 ) ;

std::atomic<long long> sum{0} ;
    parallel_for( vec.begin(), vec.end(), [&sum]( int& v ) { v += 2 ; sum += v ; } ) ;

std::cout << sum << '\n'
              << N * (N+1)/2 + N*2 << '\n' ; // should be equal to sum
}

Show compiler warnings [ + ] Compiler args [ + ] Show input
Compilation time: 3,28 sec, absolute running time: 0,22 sec, absolute service time: 3,52 sec