C supports the data types available in most CPUs, as well as some extended types. Table 17.1, “C Data Types” outlines the standard data types available in C.

C's int, long, and long long types vary in size from one platform to another. Because of this, we have to make pessimistic assumptions about the range of these types in order to write portable code. We must assume that an int has the range of a 16-bit integer and the memory requirements of a 32-bit integer, because it will on some systems. NEVER ASSUME THAT YOUR CODE WILL ONLY BE USED ON THE CPU AND OPERATING SYSTEM WHERE YOU WROTE IT. Likewise, we have to assume a long has the range of a 32-bit integer and the memory requirements of a 64-bit integer.

On 16-bit processors, int is typically 16 bits, while long is 32 bits, requiring multiple precision arithmetic. On 32-bit machines, both int and long are typically 32 bits. On 64-bit machines, int is usually 32 bits while long is usually 64 bits.

To summarize, an int could be either 16 or 32 bits, depending on where your code is compiled. A long could be wither 32 or 64 bits, depending where your code is compiled. Use int or long only if either possible size is acceptable in terms of range and memory use.

The size of int and long vary across CPUs for the sake of speed. The int type never requires multiple precision arithmetic, except perhaps on very small 8-bit microcontrollers, where an int may be 16 bits. Hence, if you want to maximize speed, and 16 bits provides enough range, and 32 bits isn't too much memory to use, then use int. Likewise, if 32 bits provides enough range, and 64 bits isn't too much memory to use, then use long.

If you need signed integers larger than +32,767 or an unsigned integer larger than 65,535, then int or unsigned int will not be big enough on 16-bit systems. Use at least a long. Always consider using unsigned before going to a larger data type. If you need values up to 50,000, but do not need negative values, then unsigned int will suffice and you do not need long, which uses more memory and may require multiple machine instructions to process.

If you want to ensure a specific size for an integer variable regardless of whether it means using multiple precision arithmetic, there are additional types defined in inttypes.h, such as int64_t and uint64_t.

#include <inttypes.h>
        
int     main(int argc, char*argv[])

{
    int32_t     myint;
    
    ...
    return 0;
}
        

If you don't need numbers larger than +32,767 and you want to limit the size of the variable to 2 bytes, then use short. This is typically only done on embedded systems with very little memory or when using very large arrays on a typical computer. Note that short values will be promoted to int in many situations, which will slow down the program.

Logical/Boolean values in C do not have a separate data type. Instead, C treats them as they are handled internally, as integers. A value of 0 represents false, and any non-zero value represents true. The standard header file stdbool.h defines a data type called bool and constants true and false. You can use these instead of integer variables and values to make programs that use Boolean variables more self-documenting.

For real numbers, use double unless saving memory is an issue. There is no significant difference in speed between float and double on modern CPUs, and double has much higher precision. Only use float for very large arrays to reduce memory use.