'C' Bitwise Operators

Relevance to 'C' of bitwise applications Syntax and expressions Example getter and setter functions Eratothene's prime number sieve

Relevance to 'C' of Bitwise Applications

'C' was designed to write system software as an alternative to assembler: compilers, kernels, device drivers, interpreters, relational database engines, virtual machines. So this language needs access to raw hardware and individual bit values. Coding designed for specific hardware design features will be non portable.

Portability Constraints

Different microprocessors organise integer and floating point data differently, e.g. big endian or little endian, two's or one's complement, location and size of exponent, sign bit and mantissa. Device drivers for different hardware implement different instruction sets.

Operator Summary

In-place operators

Inplace versions of operators exist which modify the LHS operand in place rather than returning the result for a seperate assignment, e.g. a >>= b performs a right shift of b bits directly on a . These work in the same manner as += , = and *= inplace operators compared to + and * .

Left and Right Shift Operators

The >> operator shifts a variable to the right and the << operator shifts a variable to the left. Zeros are shifted into vacated bits, but with signed data types, what happens with sign bits is platform dependant. The number of bit positions these operators shift the value on their left is specified on the right of the operator. Uses include fast multiplication or division of integers by integer powers of 2, e.g. 2,4,8,16 etc.

Left and right shift example

#include <stdio.h> int main(void){ unsigned int a=16; printf("%d\t",a>>3); /* prints 16 divided by 8 */ printf("%d\n",a<<3); /* prints 16 multiplied by 8 */ return 0; } output: 2 128

Bitwise AND and inclusive OR

Single & and | operators (bitwise AND and OR) work differently from logical AND and OR ( && and || ). You can think of the logical operators as returning a single 1 for true, and 0 for false. The purpose of the & and | bitwise operators is to return a resulting set of output 1s and 0s based on the boolean AND or OR operations between corresponding bits of the input.

Bitwise AND/OR example

#include <stdio.h> int main(void){ unsigned char a='\x00',b='\xff',c; c='\x50' | '\x07'; /* 01010000 | 00000111 */ printf("hex 50 | 07 is %x\n",c); c='\x73' & '\x37'; /* 01110011 & 00110111 */ printf("hex 73 & 37 is %x\n",c); return 0; } Output: hex 50 | 07 is 57 hex 73 & 37 is 33

Bitwise exclusive OR operator

Symbol: ^ For each bit of output, this output is a 1 if corresponding bits of input are different, and the output is a 0 if the input bits are the same.

One's complement operator

Symbol: ~ This is a unary operator in the sense that it works on a single input value. The bit pattern output is the opposite of the bit pattern input with input 1s becoming output 0s and input 0s becoming output 1s.

Setting a particular bit within a byte

Prototype used:
void setbitn(unsigned char *cp, int bitpos, int value); /* setbitn sets bit position 0 7 of cp to 0 or 1 */

The byte cp (assuming chars are 1 byte wide) is passed by reference.

Setting a particular bit within a byte

void setbitn(unsigned char *cp,int bitpos,int value){ /* setbitn sets bit position 0 7 of cp to value 0 or 1 */ unsigned char template=(unsigned char)1; /* first make template containing just the bit to set */ template<<=bitpos; if(value) /* true if value is 1 false for 0. Bitwise OR sets templated bit in cp to 1 whatever its current value, leave other bits unchanged */ *cp=*cp | template; else /* Invert template 1s and 0s. Use bitwise AND to force templated bit in cp to 0 * and leave all other bits in cp unchanged */ *cp=*cp & ( ~ template); }

Getting a particular bit within a byte

To return the value of a particular bit within a byte without changing the original, the following prototype was used:
int getbitn(unsigned char c, int bitpos); /* getbitn gets bit position 0 7 of c, returns 0 or 1 */

Call by value is used for the byte concerned, so this can be changed within the function, but as this is a copy the original byte won't be changed.

Getting a particular bit within a byte

int getbitn(unsigned char c, int bitpos){ /*getbitn gets bit position 0 7 of c, returns value 0 or 1. This function writes to c, but as we are using pass by value, this won't affect the original calling copy. */ unsigned char template=(unsigned char)1; /* make template containing just the bit to get */ template<<=bitpos; c&=template; /* if relevant bit set then c is assigned non null, otherwise c is assigned null (all zeros) */ if(c) return 1; else return 0; }

Simulating a Boolean Array

'C' doesn't have a Boolean type. We can store 0s and 1s in chars or ints one bit per variable, but this wastes space. To access individual bits from a large memoryefficient set of bits we can use a char array to allocate the storage, and then, once we have identified the correct char (byte) we can use our getbitn and setbitn functions to set and get individual bits. We can't use array notation for this, but we can design accessor and mutator functions which can be called in a similar manner to array item read ing and writing. Let's define a constant: MAXBYTES for the purpose of al locating memory to our string array, e.g. #define MAXBYTES 1000000 We can allocate the memory needed for the bit array: char bitarray[MAXBYTES]; This gives storage of MAXBYTES*8 bits, and we can use the constant within the accessor and mutator functions to avoid a buffer overrun.

getbit and setbit prototypes

void setbit(unsigned char *sp, size_t bitpos, int value); /* setbit sets bit position 0 ((MAXBYTES * 8) 1) to value 0 or 1 */ int getbit(unsigned char *sp, size_t bitpos); /* getbit gets bit position 0 ((MAXBYTES * 8) 1), returns 0 or 1 */

Parameter sp is passed the address of the array storing the bits, e.g. bitarray.

setbit function
void setbit(unsigned char *sp, size_t bitpos, int value){ /* setbit sets bit position 0 ((MAXBYTES * 8) 1) of string sp * to value 0 or 1 */ size_t byteno; int bit07; /* will store a value from 0 7 indicating bit in byte */ byteno=bitpos/8; /* floor division to get byte number in string */ if(byteno >= MAXBYTES){ fprintf(stderr,"setbit: buffer overflow trapped\n"); exit(1); /* needs #include <stdlib.h> */ } bit07=(int)bitpos%8; /* remainder value 0 7 */ setbitn(sp+byteno,bit07,value); }

getbit function
int getbit(unsigned char *sp, size_t bitpos){ /* getbit gets bit position 0 ((MAXBYTES * 8) 1) of sp, * returns 0 or 1 */ size_t byteno; int bit07; byteno=bitpos/8; /* floor division to get byte number in string */ if(byteno >= MAXBYTES){ fprintf(stderr,"getbit: read beyond end of buffer \n"); fprintf(stderr,"bitpos: %d\n",bitpos); exit(1); } bit07=(int)bitpos%8; /* remainder value 0 7 */ return getbitn(*(sp+byteno),bit07); }

Eratothene's prime number sieve

Eratothene's sieve is an algorithm we can use quickly to obtain and store all small prime num bers, e.g. less than 230 .To store these efficiently we need to use a bit array, so that if the bit for the position in the array is a 0 the number for that pos ition is prime and if the bit for this position is a 1 the number for that position is not prime. When we have all the small prime numbers, we can use these to test whether much larger randomly generated numbers are prime so they can be used for generating cryptographic keys.

Eratothene's prime number sieve

Positions in this array all start at 0 (meaning the number or index of the position could be prime) except for positions 0 and 1 (which are not prime numbers. Starting with 2 this number is multiplied by 2, 3, 4 etc. to strike out the even numbers starting 4, 6, 8 etc. up to the highest posi tion in the boolean array. These bits are set to 1s, mean ing they can't be prime. The next 0 in the array is then found, which is the next prime number 3, and this number is multiplied by all higher numbers, and all multiples are set to 1. This process continues until the first prime great er than the square root of the highest bit position is found. Bitwise operators are needed to implement Eratosthenes Sieve algorithm without wasting memory.

Eratothenes Sieve main function

int main(void){ /* from eratosthenes.c , Richard Kay, Sept 2005 */ int i,j,nextprime=2,k; unsigned char a[MAXBYTES]; /* allocate memory for bit array */ for(i=0;i<MAXBYTES;i++) a[i]=0x00; /* initialise all bits to 0s */ setbit(a,0,1); setbit(a,1,1); /* seed 0 and 1 as not prime numbers */ printprime(2); while(nextprime+1<MAXBYTES*8){ k=nextprime; /* mark multiples of nextprime as not being possibly prime */ while(nextprime*k<MAXBYTES*8){ setbit(a,nextprime*k,1); k++; } /* find nextprime by skipping nonprime bits marked 1 */ while(nextprime+1<MAXBYTES*8 && getbit(a,++nextprime)); printprime(nextprime); } return 0; }

Eratothenes Sieve printprime function

void printprime(int prime){ /* prints a prime number in next 8 columns */ static int numfound=0; /* static so only initialises at compile time, and previous values are remembered in subsequent calls */ if(numfound%8==0) printf("\n"); /* start next row */ if(prime+1<MAXBYTES*8) printf("%d\t",prime); numfound++; } The above program is quite simple, but we didn't need to store whether multiples of 2 were prime or not. A minor improve ment to our data structure design will halve the memory.

Further reading on primes

The sieve of Atkin is a faster version of Eratothene's sieve http://en.wikipedia.org/wiki/Sieve_of_Atkin Small primes can be used to test whether a random number is a much larger prime using the Rabin Miller primality test. http://en.wikipedia.org/wiki/Primality_test