Reputation: 11926
I am using anonymous structs in unions to get a%b quickly.
Do you know any other way to get a%b without using powers of 2 for b.
include list:
#include<stdio.h>
#include<stdlib.h>
#include<time.h>
and declaration of unions:
//C99
//my first program to test rand() for any periodicity
union //for array indexing without modulus operator
{
unsigned int counter; //32 bit on my computer
struct
{
unsigned int lsb:16; //16 bit
unsigned int msb:16; //16 bit
};
struct
{
unsigned int bit[32];
};
} modulo;
union // for rand()%256
{
unsigned int randoom; //from rand()
struct
{
unsigned int lsb:5;//equivalent to rand()%32 without any calculations
unsigned int msb:27;
};
}random_modulus;
and here the main function:
int main()
{
srand(time(0));
modulo.counter=0;//init of for-loop counter
// i am takin (counter%65536) for my array index which is modulus.lsb
unsigned int array_1[65536];
float array_mean=0,delta_square=0;
clock_t clock_refe;
//taking counter%65536 using lsb (2x faster)
clock_refe=clock();
for(;modulo.counter<1000000000;modulo.counter++)
{
// i need accumulated randoms for later use for some std. dev. thing.
random_modulus.randoom=rand();
array_1[modulo.lsb]+=random_modulus.lsb;
}
//getting average clock cycles
for(int i=0;i<65536;i++)
{
array_mean+=array_1[i];
}
array_mean/=65536.0f;
//getting square of deltas
float array_2[65536];
for(int i=0;i<65536;i++)
{
array_2[i]=(array_1[i]-array_mean)*(array_1[i]-array_mean);
}
//horizontal histogram for resoluton of 20 elements
for(int i=0;i<65536;i+=(65536)/20)
{
for(int j=0;j<(array_2[i]*0.01);j++)
{
printf("*");
}
printf("\n");
}
//calculations continue .....
return 0;
}
Pre-calculated values in some array maybe? But if i use the % calc part for only once, this is same. Can you give some book references about bitwise operations manuals?
Upvotes: 0
Views: 202
Reputation: 5165
Bitwise operations are the most portable method in this case. Here's some example functions; they are written for readability, but can be easily made faster:
int lsb(int input) {
int mask = 0x00FF; // 0x00FF is 0000000011111111 in binary
return input & mask;
}
int msb(int input) {
int mask = 0xFF00; // 0xFF00 is 1111111100000000 in binary
return (input & mask) >> 8;
}
In general, mask the bits you want with &
, and then align them to the right with >>
. K&R 2nd edition (The C Programming Language by Brian Kernighan & Dennis Ritchie) has information on just about every C topic, including bitmasks.
If you want a % b
where b
is not a power of 2, the native %
operator is the fastest way (in modern compilers it's as fast as the bitwise operations, even when b
is a power of two). Bitwise operations are useful in other contexts, though.
Upvotes: 1