Padding is actually a compiler optimization technique, which
fill up the area between to member to make multiples of 4
bytes (32 bits). Only combination chars/ shorts int/ both ca
e able to create a 32 bytes. Compiler actually will skip to
check the size the variable and fetch the whole 4 bytes in
data fetch operation. this will provide the aligned data to
MP in Single CPU cycle.
There is a Pragma directive, which override the compiler and
results the structure size with equal to size of variables.
 
Example:

Struct item {
              int item;
              char type;
};

Sizeof (struct item) = 8 bytes

Struct item {
              int item;
              char type[2];
              short int value;
};
Sizeof (struct item) = 8 bytes

Struct item {
              int item;
              char type[3];
              short int value;
};
Sizeof (struct item) = 12 bytes
one byte padded after "type" and 2 byes padded after value.



Bit-Wise: As far as I remember, it maintains a ARRAY of
int/char/bits equal to the number of elements defined inside
. The bit sets if some data is initialized with some value.
I am not sure on this.