Re: [boost] Detection of 64bit using macro

Subject: Re: [boost] Detection of 64bit using macro
From: Andrey Semashev (andrey.semashev_at_[hidden])
Date: 2010-08-14 06:32:13

On 08/13/2010 11:44 PM, Kim Kuen Tang wrote:
>
> Hi all,
>
> the question might be not directly related to boost. But i just want to
> be sure that i am not reventing the wheel.
> I am searching for a macro that is able to detect that a system is 32bit
> or 64bit no matter what compiler is used.
> Is there something similar already implemented in boost?

There is no standard macro to detect this, so every solution will be 
platform-specific. For instance, I think, GCC defines some macro that 
has value of sizeof(void*). On platforms with uintptr_t you can test for 
UINTPTR_MAX macro value. On linux there is __WORDSIZE macro that has 
value of number of bits in the native integer type (that is, 32 or 64). 
Some compilers may define their own specific macros:

<http://predef.sourceforge.net/prearch.html>

> Perhaps some background about this feature.
>
> Template specialization for std::size_t is not needed, when the
> specialization already exists for unsigned int. But on a 64bit system
> this is not true.
> So you need explicitly added the specialization whenever a 64bit
> compiler is used.
>
> See the following example.
>
> # include <cstddef>
>
> template<typename T>
> struct Null;
>
> template<> struct Null<unsigned int> {};
> template<> struct Null<std::size_t> {}; // compile only on a 64bit compiler
>
> int main()
> {
> return 0;
> };
>
> Thx for any comments, help or suggestions

In this case I would do something like this:

   template< typename T >
   struct Null:
     public NullImpl< T, sizeof(T) >
   {
   };

and then specialize NullImpl on the size of the type, rather than the 
type itself.