[Bug tree-optimization/122101] Large performance difference between GCC and Clang at -O3

[arash at partow dot net via Gcc-bugs]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=122101

--- Comment #3 from Arash Partow <arash at partow dot net> ---
@AndrewPinski - I think you're right. Removing the integer mod operation,
brings both in to the same performance range. 

This is the modified version:

------ code ------

inline bool is_prime(std::uint32_t i)
{
    if (i == 2) return true;
    if ((i == 1) || (i % 2 == 0)) return false;

    const auto f = static_cast<double>(i);
    const auto c = std::ceil(std::sqrt(f));

    for (double j = 3.0; j <= c; j += 2.0)
    {
        const double g = f / j;
        if ((g - std::trunc(g)) == 0.0) return false;
    }

    return true;
}

------ code ------

Note: using float or double doesn't make a difference.