Packing Bits with Bit Fields

Packing Bits C lets you pack bits tightly without requiring a lot of anding, and shifting. Lets assume we want to pack 4 values into one byte. These are a,b,c and d, defined like this.

 ddddcbaa

If we call this byte x then the value of d is

 d = (x >> 4) &0xf;

Which is a bit convoluted. What if we could just access the value of d and let the compiler do all the messy bit manipulation? Well we can but we have to define the bits using a bit field. To define our byte, we use a struct as follows.

 struct flagtype
 {
 unsigned char a: 2;
 unsigned char b: 1;
 unsigned char c : 1;
 unsigned char d : 4;
 };
 
 struct flagtype f;
 
 int x;
 x=f.d; // fetch value

The values such as : 4 define how big each field is in bits. The type makes it a bit more complicated as it affects the position of the bits in the struct. Change b to an int and flagtype increases in size from 1 byte to 12! Change a as as well to an int though and flagtype shrinks to 8 bytes! This suggests that it is organizing the bits to start on a 4 byte boundary, presumably for optimal access speed. My guess is that it is something like this when a is a char and b an int:

 00000000 
 00000000 
 00000000 
 000ddddc 
 00000000 
 00000000 
 00000000 
 0000000b
 00000000 
 00000000 
 00000000 
 000000aa

Change a to an int as well and it all fits in 8 bytes like this:

 00000000 
 00000000 
 00000000 
 000ddddc 
 00000000 
 00000000 
 00000000 
 00000baa

The actual layout implementation is probably compiler dependant. On Visual C++ 6, flagtype as declared is exactly one byte in size. Example 6 uses a union to access a variable of type flagtype. By assigning values to a,b,c etc, you can identify the layout of the bits.


ex:

#include 


struct flagtype
{
 unsigned char a:  2;
 unsigned int  b:  1;
 unsigned char c : 1;
 unsigned char d : 4;
};

union utype {
 struct flagtype f;
 unsigned char  byte;
};


int main(int argc, char* argv[])
{
 union utype u;
 int x;
 u.byte=0;
 printf("Size of u = %d\n",sizeof(u));
 u.f.a=3;
 u.f.b=0;
 u.f.c=0;
 u.f.d=15;
 printf("Byte = %d",u.byte);
 x=u.f.d;
 return 0;}
}