How do I reinterpret data through a different type? (type punning confusion)

Question

#include 

int main(int argc, char * argv[])
{
    int a = 0x3f800000;

    std::cout << a << std::endl;

    static_assert(sizeof(float)         


        

Answer 1

As you've found out, reinterpret_cast can't be used for type punning

• Is reinterpret_cast type punning actually undefined behavior?
• Is reinterpret_cast mostly useless?

Since C++20 you'll have a safe way for type punning via std::bit_cast

uint32_t bit_pattern = 0x3f800000U;
constexpr auto f = std::bit_cast(bit_pattern);

At the moment std::bit_cast is only supported by MSVC

While waiting for others to implement that, if you're using Clang you can try __builtin_bit_cast. Just cast it like this

float f = __builtin_bit_cast(float, bit_pattern);

See demo on Godbolt

---

In other compilers or older C++ standards the only way you can do is via a memcpy

However many compilers have implementation-specific way to do type punning, or implementation-specific behavior regarding type punning. For example in GCC you can use __attribute__((__may_alias__))

union Float
{
    float __attribute__((__may_alias__)) f;
    uint32_t __attribute__((__may_alias__)) u;
};

uint32_t getFloatBits(float v)
{
    Float F;
    F.f = v;
    return F.u;
}

ICC and Clang also support that attribute. See demo