overflow – Unsigned integer underflow optimization

I’m having some problem with the following issue:

#include <iostream>

using namespace std;

int main()
{
    uint8_t l = 200, r = 2, c = 199;

    bool res1 = (l - c) < (r - c);
    bool res2 = uint8_t(l - c) < uint8_t(r - c);
    
    uint8_t ll = l - c;
    uint8_t rr = r - c;
    bool res3 = ll < rr;

    std::cout << std::boolalpha << res1 << std::endl;
    std::cout << std::boolalpha << res2 << std::endl;
    std::cout << std::boolalpha << res3 << std::endl;
    
    return 0;
}

Compiling int with g++ and clang++ I got

 false
 true
 true

Now the whole point is that I would like to exploit the unsigned underflow , i.e. I expect the second and third answer. I imagine that in the first case, the compiler elide on both side c. I would like to know whether there is a robust and compiler-independent way to handle this (without explicitly checking for underflow, which can lower the performance).