Pi problem with OMP

#include <iostream>
#include <algorithm>
#include <limits>

#include <omp.h>

double sum(const uint32_t &id, const uint64_t &num_steps, const uint8_t &threads_count)
{
    double x, sum{0.};
    const double step = 1. / num_steps;
    const uint32_t chunk_size = num_steps / threads_count;
    for (uint_fast32_t i(id * chunk_size), _n((id + 1) * chunk_size); i < _n; ++i)
    {
        x = (i + 0.5) * step;
        sum += 4. / (1 + x * x);
    }
    return sum;
}

int main()
{
    const uint64_t num_steps = 1000 * 1000 * 1000;
    const uint8_t threads_count = 4;
    double res[threads_count];
#pragma omp parallel num_threads(threads_count)
    {
        uint32_t id = ::omp_get_thread_num();
        res[id] = sum(id, num_steps, threads_count);
    }
    ::std::cout.precision(::std::numeric_limits<double>::max_exponent10);
    ::std::cout << ::std::fixed << ::std::accumulate(res, res + 4, 0.) / num_steps << ::std::endl;
}