How to output IEEE-754 format integer as a float

Question

I have a unsigned long integer value which represents a float using IEEE-754 format. What is the quickest way of printing it out as a float in C++?

I know one way, but am wondering if there is a convenient utility in C++ that would be better.

Example of the way that I know is:

union
{
    unsigned long ul;
    float f;
} u;

u.ul = 1084227584; // in HEX, this is 0x40A00000

cout << "float value is: " << u.f << endl;

(This prints out "float value is: 5" )

Answer 1

The union method you suggested is the usual route that most people would take. However, it's technically undefined behavior in C/C++ to read a different member from a union than the one that was most recently written. Despite this, though, it's well-supported among pretty much all compilers.

Casting pointers, as Jon Skeet suggested, is a bad idea -- that violates the strict aliasing rules of C. An aggressive optimizing compiler will produce incorrect code for this, since it assumes that pointers of types unsigned long * and float * will never alias each other.

The most correct way, in terms of standards compliance (that is, not invoking undefined behavior), is to cast through a char*, since the strict aliasing rules permit a char* pointer to alias a pointer of any other type:

unsigned long ul = 0x40A00000;
float f;
char *pul = (char *)&ul;  // ok, char* can alias any type
char *pf = (char *)&f;    // ok, char* can alias any type
memcpy(pf, pul, sizeof(float));

Though honestly, I would just go with the union method. From the cellperformance.com link above:

It is an extremely common idiom and is well-supported by all major compilers. As a practical matter, reading and writing to any member of a union, in any order, is acceptable practice.

Answer 2

It will only work on 32-bit machines or machines where sizeof(long) == sizeof(float). On modern machines you might want to use int instead ... and you really have to take care if you're performing this trick.