Question

Can anybody tell me how to make this C code work in C++ please?

uint64_t flv_dbl2int( double value )
{
    return (union {double f; uint64_t i;}){value}.i;    
}

I’m not precisely sure what is going on, is it bit shift?

Answer 1

As far as I can tell, the only legal strategy for type punning in C++ is through memcpy(), ie

uint64_t flv_dbl2int(double fvalue)
{
    uint64_t ivalue;
    memcpy(&ivalue, &fvalue, sizeof ivalue);
    return ivalue;
}

The straight-forward solution using reinterpret_cast

uint64_t flv_dbl2int(double value)
{
    return reinterpret_cast<uint64_t&>(value);
}

is actually undefined behaviour, same as using C-style pointer casts.

Using unions (as the original solution does), while legal in C, is probably illegal in C++. In practice, all of these solutions are likely to work, and at least gcc-4.5.3 and clang-3.0 will produce identical code on x86 at -O1.