Below is code that will compile when INT_MAX is equal to 232. It will generate a compiler error when INT_MAX is not equal to 232.

#include <iostream>
#include <climits>

namespace NBitCheck
    template<bool> struct CTAssert;
    template<> struct CTAssert<true> {};

    constexpr bool intbitwidthcheck=INT_MAX-(2^32)+1;
    CTAssert<intbitwidthcheck> a;

int main(void) {

  return 0;

In C++, sizeof returns the size of a number relative to the size of char.

So, sizeof(int) should give you how many chars each int takes.

Next, we need to know how many bits each char has. CHAR_BIT has that information.

So try this:

constexpr size_t bits_per_int = sizeof(int) * CHAR_BIT;
constexpr bool intbitwidthcheck = (bits_per_int == 32);

Frankly, I believe this way is more readable.

Not to mention, can one reasonably expect a 32-bits signed int to contain values up to 2^32 - 1? Shouldn't that be the UINT_MAX? INT_MAX should be up to 2^31 - 1.

  • \$\begingroup\$ One may expect (even though I never saw such architecture) that INT_MAX == UINT_MAX as a border case. What the standard ensures is that The range of non-negative values of a signed integer type is a subrange of the corresponding unsigned integer type, and the value representation of each corresponding signed/unsigned type shall be the same. \$\endgroup\$ – Massimiliano Jan 26 '14 at 10:26
  • \$\begingroup\$ horrible C-style hack. Why not use the C++ features explicitly provided for such purposes? (std::numeric_limits<>) \$\endgroup\$ – Walter Jan 30 '14 at 21:11
  • \$\begingroup\$ @Walter numeric_limits doesn't tell how many bits an int has. If instead of checking how many bits there are you really want the actual representable limits, use Massimiliano's approach instead. \$\endgroup\$ – luiscubal Jan 31 '14 at 2:17

To check if the fundamental type int is 32 bits wide I would do the following:


static_assert( std::numeric_limits<int>::max() == std::numeric_limits<int32_t>::max() , "int is not 32 bits wide");

which should be (to the best of my understanding) safer than what is proposed by @luiscubal.

The point I would raise with the accepted answer is that not all the bits in the object representation of an int may be used to represent a numeric value.

In fact, even though in the C++ standard (section 3.9) the issue is not clearly stated as in the C standard (section, I think the possibility of having padding bits in an integer type representation is allowed for compatibility reasons.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.