I wrote the following code to list all the prime numbers upto 2 billion using Sieve's method. I used bitmasking for flagging purpose. While I am able to get the prime numbers correctly, a few primes in the beginning are missing every time. Please help me find the bug in the program.
#include <stdio.h>
#include <stdlib.h>
#include <math.h>
#include <stdbool.h>
#define MAX 2000000000
char* listPrimes(){
int block = sqrt(MAX);
char* mark = calloc((MAX/8),sizeof(char));
int i = 2;
int j;
char mask[8];
for(j=0;j<8;j++)
mask[j] = 0;
mask[7] = 1;
mask[6] |= mask[7] << 1;
mask[5] |= mask[7] << 2;
mask[4] |= mask[7] << 3;
mask[3] |= mask[7] << 4;
mask[2] |= mask[7] << 5;
mask[1] |= mask[7] << 6;
mask[0] |= mask[7] << 7;
for(j=0;j<8;j++)
printf("%d ",mask[j]);
mark[0] |= mask[0];
mark[0] |= mask[1];
while (i < block){
for (j = 2; i*j <= block; j++)
mark[(i*j) / 8] |= mask[((i*j) % 8 )];
i++;
}
printf("\n");
printf("The block size is\t:\t%d\n",block);
j = 2;
while(j<=block){
if((mark[j / 8] & mask[j]) == 0 ){
for(i = 2;i <= MAX; i++){
if((i%j) == 0){
mark[i / 8] |= mask[(i % 8)];
}
}
}
while((mark[++j / 8] & mask[j % 8]) != 0);
}
for(j=0;j<=MAX;j++)
if((mark[j / 8] & mask[(j % 8)]) == 0)
printf("%d\n", ((8*(j / 8)) + (j % 8)));
return mark;
}
int main(int argc,char* argv[]){
listPrimes();
return 0;
}
As ArunMK said, in the second while loop you mark the prime j itself as a multiple of j. And as Lee Meador said, you need to take the modulus of j modulo 8 for the mask index, otherwise you access out of bounds and invoke undefined behaviour.
A further point where you invoke undefined behaviour is
while((mark[++j / 8] & mask[j % 8]) != 0);
where you use and modify j without intervening sequence point. You can avoid that by writing
do {
++j;
}while((mark[j/8] & mask[j%8]) != 0);
or, if you insist on a while loop with empty body
while(++j, (mark[j/8] & mask[j%8]) != 0);
you can use the comma operator.
More undefined behaviour by accessing mark[MAX/8] which is not allocated in
for(i = 2;i <= MAX; i++){
and
for(j=0;j<=MAX;j++)
Also, if char is signed and eight bits wide,
mask[0] |= mask[7] << 7;
is implementation-defined (and may raise an implementation-defined signal) since the result of
mask[0] | (mask[7] << 7)
(the int 128) is not representable as a char.
But why are you dividing each number by all primes not exceeding the square root of the bound in the second while loop?
for(i = 2;i <= MAX; i++){
if((i%j) == 0){
That makes your algorithm not a Sieve of Eratosthenes, but a trial division.
Why don't you use the technique from the first while loop there too? (And then, why two loops at all?)
while (i <= block){
if ((mark[i/8] & mask[i%8]) == 0) {
for (j = 2; i*j < MAX; j++) {
mark[(i*j) / 8] |= mask[((i*j) % 8 )];
}
}
i++;
}
would not overflow (for the given value of MAX, if that is representable as an int), and produce the correct output orders of magnitude faster.
Change the middle loop to add the modulo:
j = 2;
while(j<=block){
if((mark[j / 8] & mask[j % 8]) == 0 ){
for(i = 2;i <= MAX; i++){
if((i%j) == 0){
mark[i / 8] |= mask[(i % 8)];
}
}
}
}
In the second while loop you are looping through i from 2 onwards and you do an if (i%j == 0). This will be true for i when it is a prime as well. You need to check for (i != j). Also the modulo as reported above. Hence it becomes:
if ((i%j == 0) {
if (i!=j)
mark[i/j] |= mask[i%j];
}
Related
Dear all C programmer:
X = 1 << N; (left shift)
how to recover N from X ?
Thanks
N in this case is the bit position where you shifted in a 1 at. Assuming that X here only got one bit set. Then to find out what number that bit position corresponds to, you have to iterate through the data and mask with bitwise AND:
for(size_t i=0; i<sizeof(X)*8; i++)
if(X & (1<<i))
printf("%d", i);
If performance is important, then you'd make a look-up table with all possible results instead.
In a while loop, keep shifting right until X==1, record how many times you have to shift right and the counter will give you N.
int var = X;
int count = 0;
while (var != 1){
var >>= 1;
count++;
}
printf("N is %d", count);
Try this (flsl from here which is available from string.h on macOS) :
int flsl(long mask)
{
int bit;
if (mask == 0) return (0);
for (bit = 1; mask != 1; bit++)
mask = (unsigned long)mask >> 1;
return (bit);
}
unsigned char binlog(long mask) { return mask ? flsl(mask) - 1 : 0; }
int x = 1 << 20;
printf("%d\n", binlog(x)); ===> 20
I'm wondering if someone know effective approach to calculate bits in specified position along array?
Assuming that OP wants to count active bits
size_t countbits(uint8_t *array, int pos, size_t size)
{
uint8_t mask = 1 << pos;
uint32_t result = 0;
while(size--)
{
result += *array++ & mask;
}
return result >> pos;
}
You can just loop the array values and test for the bits with a bitwise and operator, like so:
int arr[] = {1,2,3,4,5};
// 1 - 001
// 2 - 010
// 3 - 011
// 4 - 100
// 5 - 101
int i, bitcount = 0;
for (i = 0; i < 5; ++i){
if (arr[i] & (1 << 2)){ //testing and counting the 3rd bit
bitcount++;
}
}
printf("%d", bitcount); //2
Note that i opted for 1 << 2 which tests for the 3rd bit from the right or the third least significant bit just to be easier to show. Now bitCount would now hold 2 which are the number of 3rd bits set to 1.
Take a look at the result in Ideone
In your case you would need to check for the 5th bit which can be represented as:
1 << 4
0x10000
16
And the 8th bit:
1 << 7
0x10000000
256
So adjusting this to your bits would give you:
int i, bitcount8 = 0, bitcount5 = 0;
for (i = 0; i < your_array_size_here; ++i){
if (arr[i] & 0x10000000){
bitcount8++;
}
if (arr[i] & 0x10000){
bitcount5++;
}
}
If you need to count many of them, then this solution isn't great and you'd be better off creating an array of bit counts, and calculating them with another for loop:
int i, j, bitcounts[8] = {0};
for (i = 0; i < your_array_size_here; ++i){
for (j = 0; j < 8; ++j){
//j will be catching each bit with the increasing shift lefts
if (arr[i] & (1 << j)){
bitcounts[j]++;
}
}
}
And in this case you would access the bit counts by their index:
printf("%d", bitcounts[2]); //2
Check this solution in Ideone as well
Let the bit position difference (e.g. 7 - 4 in this case) be diff.
If 2diff > n, then code can add both bits at the same time.
void count(const uint8_t *Array, size_t n, int *bit7sum, int *bit4sum) {
unsigned sum = 0;
unsigned mask = 0x90;
while (n > 0) {
n--;
sum += Array[n] & mask;
}
*bit7sum = sum >> 7;
*bit4sum = (sum >> 4) & 0x07;
}
If the processor has a fast multiply and n is still not too large, like n < pow(2,14) in this case. (Or n < pow(2,8) in the general case)
void count2(const uint8_t *Array, size_t n, int *bit7sum, int *bit4sum) {
// assume 32 bit or wider unsigned
unsigned sum = 0;
unsigned mask1 = 0x90;
unsigned m = 1 + (1u << 11); // to move bit 7 to the bit 18 place
unsigned mask2 = (1u << 18) | (1u << 4);
while (n > 0) {
n--;
sum += ((Array[n] & mask1)*m) & mask2;
}
*bit7sum = sum >> 18;
*bit4sum = ((1u << 18) - 1) & sum) >> 4);
}
Algorithm: code is using a mask, multiply, mask to separate the 2 bits. The lower bit remains in it low position while the upper bit is shifted to the upper bits. Then a parallel add occurs.
The loop avoids any branching aside from the loop itself. This can make for fast code. YMMV.
With even larger n, break it down into multiple calls to count2()
I have a program that requires me to find primes up till 10**10-1 (10,000,000,000). I wrote a Sieve of Eratosthenes to do this, and it worked very well (and accurately) as high as 10**9 (1,000,000,000). I confirmed its accuracy by having it count the number of primes it found, and it matched the value of 50,847,534 on the chart found here. I used unsigned int as the storage type and it successfully found all the primes in approximately 30 seconds.
However, 10**10 requires that I use a larger storage type: long long int. Once I switched to this, the program is running signifigantly slower (its been 3 hours plus and its still working). Here is the relevant code:
typedef unsigned long long ul_long;
typedef unsigned int u_int;
ul_long max = 10000000000;
u_int blocks = 1250000000;
char memField[1250000000];
char mapBit(char place) { //convert 0->0x80, 1->0x40, 2->0x20, and so on
return 0x80 >> (place);
}
for (u_int i = 2; i*i < max; i++) {
if (memField[i / 8] & activeBit) { //Use correct memory block
for (ul_long n = 2 * i; n < max; n += i) {
char secondaryBit = mapBit(n % 8); //Determine bit position of n
u_int activeByte = n / 8; //Determine correct memory block
if (n < 8) { //Manual override memory block and bit for first block
secondaryBit = mapBit(n);
activeByte = 0;
}
memField[activeByte] &= ~(secondaryBit); //Set the flag to false
}
}
activeBit = activeBit >> 1; //Check the next
if (activeBit == 0x00) activeBit = 0x80;
}
I figure that since 10**10 is 10x larger then 10**9 it should take 10 times the amount of time. Where is the flaw in this? Why did changing to long long cause such significant performance issues and how can I fix this? I recognize that the numbers get larger, so it should be somewhat slower, but only towards the end. Is there something I'm missing.
Note: I realize long int should technically be large enough but my limits.h says it isn't even though I'm compiling 64 bit. Thats why I use long long int in case anyone was wondering. Also, keep in mind, I have no computer science training, just a hobbyist.
edit: just ran it in "Release" as x86-64 with some of the debug statements suggested. I got the following output:
looks like I hit the u_int bound. I don't know why i is getting that large.
Your program has an infinite loop in for (u_int i = 2; i*i < max; i++). i is an unsigned int so i*i wraps at 32-bit and is always less than max. Make i an ul_long.
Note that you should use simpler bit pattern from 1 to 0x80 for bit 0 to 7.
Here is a complete version:
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
typedef unsigned long long ul_long;
typedef unsigned int u_int;
#define TESTBIT(a, bit) (a[(bit) / 8] & (1 << ((bit) & 7)))
#define CLEARBIT(a, bit) (a[(bit) / 8] &= ~(1 << ((bit) & 7)))
ul_long count_primes(ul_long max) {
size_t blocks = (max + 7) / 8;
unsigned char *memField = malloc(blocks);
if (memField == NULL) {
printf("cannot allocate memory for %llu bytes\n",
(unsigned long long)blocks);
return 0;
}
memset(memField, 255, blocks);
CLEARBIT(memField, 0); // 0 is not prime
CLEARBIT(memField, 1); // 1 is not prime
// clear bits after max
for (ul_long i = max + 1; i < blocks * 8ULL; i++) {
CLEARBIT(memField, i);
}
for (ul_long i = 2; i * i < max; i++) {
if (TESTBIT(memField, i)) { //Check if i is prime
for (ul_long n = 2 * i; n < max; n += i) {
CLEARBIT(memField, n); //Reset all multiples of i
}
}
}
unsigned int bitCount[256];
for (int i = 0; i < 256; i++) {
bitCount[i] = (((i >> 0) & 1) + ((i >> 1) & 1) +
((i >> 2) & 1) + ((i >> 3) & 1) +
((i >> 4) & 1) + ((i >> 5) & 1) +
((i >> 6) & 1) + ((i >> 7) & 1));
}
ul_long count = 0;
for (size_t i = 0; i < blocks; i++) {
count += bitCount[memField[i]];
}
printf("count of primes up to %llu: %llu\n", max, count);
free(memField);
return count;
}
int main(int argc, char *argv[]) {
if (argc > 1) {
for (int i = 1; i < argc; i++) {
count_primes(strtoull(argv[i], NULL, 0));
}
} else {
count_primes(10000000000);
}
return 0;
}
It completes in 10 seconds for 10^9 and 131 seconds for 10^10:
count of primes up to 1000000000: 50847534
count of primes up to 10000000000: 455052511
My professor asked me to do a program to test the Goldbach conjecture. I am wondering if I should consider 1 as a prime. This is my code that prints the first combination of prime numbers:
#include <stdio.h>
#include <stdlib.h>
#include <math.h>
int main()
{
int n, i, j, k, l;
char prime, prime1;
do //check if n is even and greater than 2
{
printf("Give me an even natural number greater than 2:\n\n>");
scanf("%d", &n);
}
while (n % 2 != 0 && n >= 2);
for (i = 1; i < n ; i++)
{
prime = 1;
for (k = 2; k < i; k++)
if (i % k == 0)
prime = 0;
if (prime)
{
for (j = 1; j < n; j++)
{
prime1 = 1;
for (l = 2; l < j; l++)
if (j % l == 0)
prime1 = 0;
if (prime1)
if (i + j == n)
{
printf("\n\n%d and %d are the first two prime numbers that add up to %d.\n", i, j, n);
return 0;
}
}
}
}
}
I checked the internet and almost everyone says that 1 is not a prime. What should i do? Keep the program as it is or change it so that it won't consider 1 as a prime? And how do I do that? :P
you can consider 1 a prime number, as Goldbach did, or not as is the more common usage, it makes almost no difference regarding the conjecture.
Considering 1 as a prime number has this effect:
there is a solution for 2: 1 + 1.
the first pair for 4 is 1 + 3 instead of 2 + 2
the first solution for higher even numbers may involve 1 if the value is a prime number plus one, but no known even number greater than 2 can only be expressed as p + 1.
Note that there are problems in your code:
you do not check the return value of scanf(), so inputing a string that is not a number will cause undefined behavior (the first time as n is uninitialized) or an infinite loop as n is no longer modified.
the test while (n % 2 != 0 && n >= 2); is incorrect: it should be:
while (n <= 2 || n % 2 != 0);
the first loop could iterate half as long with a test i <= n / 2
the second loop could iterate much less with a test k * k <= i
you could exit the second loop when you detect that i is not prime
there is no need for a third loop, you just need to test if n - i is prime
the same improvements are possible for the second primary test, better move this to a separate function.
you should have a message and return statement for the remote possibility that you find a counter-example to the Goldbach conjecture ;-)
Here is an improved version:
#include <stdio.h>
#define PRIME_MASK ((1ULL << 2) | (1ULL << 3) | (1ULL << 5) | (1ULL << 7) |\
(1ULL << 11) | (1ULL << 13) | (1ULL << 17) | (1ULL << 19) | \
(1ULL << 23) | (1ULL << 29) | (1ULL << 31) | (1ULL << 37) | \
(1ULL << 41) | (1ULL << 43) | (1ULL << 47) | (1ULL << 53) | \
(1ULL << 59) | (1ULL << 61))
int isprime(unsigned long long n) {
if (n <= 63)
return (PRIME_MASK >> n) & 1;
if (n % 2 == 0)
return 0;
for (unsigned long long k = 3; k * k <= n; k += 2) {
if (n % k == 0)
return 0;
}
return 1;
}
int main(void) {
unsigned long long n, i;
int r;
for (;;) {
printf("Give me an even natural number greater than 2:\n>");
r = scanf("%llu", &n);
if (r == 1) {
if (n % 2 == 0 && n > 2)
break;
} else
if (r == EOF) { /* premature end of file */
return 1;
} else {
scanf("%*[^\n]%*c"); /* flush pending line */
}
}
#ifdef ONE_IS_PRIME
i = 1; /* start this loop at 1 if you want to assume 1 is prime */
#else
i = (n == 4) ? 2 : 3;
#endif
for (; i <= n / 2; i += 2) {
if (isprime(i) && isprime(n - i)) {
printf("%llu = %llu + %llu\n", n, i, n - i);
return 0;
}
}
printf("Goldbach was wrong!\n"
" %llu cannot be written as the sum of two primes\n", n);
return 0;
}
YOU CAN Consider 1 to be a prime as Goldbach too considered it to be prime in his letter to Leonhard Euler.But that was the time when 1 was considered to be prime.Later it was abandoned and hence this one is Goldbach's third revised conjecture.Also,since today we consider 1 to be neither prime nor composite,even if you don't consider 1 to be a prime number,the conjecture still holds true,is well-verified upto 4*10^18 (re-verified upto 4*10^17).
As far as you deal with the Professor,you should better ask him what does he want.
I want to extract bits of a decimal number.
For example, 7 is binary 0111, and I want to get 0 1 1 1 all bits stored in bool. How can I do so?
OK, a loop is not a good option, can I do something else for this?
If you want the k-th bit of n, then do
(n & ( 1 << k )) >> k
Here we create a mask, apply the mask to n, and then right shift the masked value to get just the bit we want. We could write it out more fully as:
int mask = 1 << k;
int masked_n = n & mask;
int thebit = masked_n >> k;
You can read more about bit-masking here.
Here is a program:
#include <stdio.h>
#include <stdlib.h>
int *get_bits(int n, int bitswanted){
int *bits = malloc(sizeof(int) * bitswanted);
int k;
for(k=0; k<bitswanted; k++){
int mask = 1 << k;
int masked_n = n & mask;
int thebit = masked_n >> k;
bits[k] = thebit;
}
return bits;
}
int main(){
int n=7;
int bitswanted = 5;
int *bits = get_bits(n, bitswanted);
printf("%d = ", n);
int i;
for(i=bitswanted-1; i>=0;i--){
printf("%d ", bits[i]);
}
printf("\n");
}
As requested, I decided to extend my comment on forefinger's answer to a full-fledged answer. Although his answer is correct, it is needlessly complex. Furthermore all current answers use signed ints to represent the values. This is dangerous, as right-shifting of negative values is implementation-defined (i.e. not portable) and left-shifting can lead to undefined behavior (see this question).
By right-shifting the desired bit into the least significant bit position, masking can be done with 1. No need to compute a new mask value for each bit.
(n >> k) & 1
As a complete program, computing (and subsequently printing) an array of single bit values:
#include <stdio.h>
#include <stdlib.h>
int main(int argc, char** argv)
{
unsigned
input = 0b0111u,
n_bits = 4u,
*bits = (unsigned*)malloc(sizeof(unsigned) * n_bits),
bit = 0;
for(bit = 0; bit < n_bits; ++bit)
bits[bit] = (input >> bit) & 1;
for(bit = n_bits; bit--;)
printf("%u", bits[bit]);
printf("\n");
free(bits);
}
Assuming that you want to calculate all bits as in this case, and not a specific one, the loop can be further changed to
for(bit = 0; bit < n_bits; ++bit, input >>= 1)
bits[bit] = input & 1;
This modifies input in place and thereby allows the use of a constant width, single-bit shift, which may be more efficient on some architectures.
Here's one way to do it—there are many others:
bool b[4];
int v = 7; // number to dissect
for (int j = 0; j < 4; ++j)
b [j] = 0 != (v & (1 << j));
It is hard to understand why use of a loop is not desired, but it is easy enough to unroll the loop:
bool b[4];
int v = 7; // number to dissect
b [0] = 0 != (v & (1 << 0));
b [1] = 0 != (v & (1 << 1));
b [2] = 0 != (v & (1 << 2));
b [3] = 0 != (v & (1 << 3));
Or evaluating constant expressions in the last four statements:
b [0] = 0 != (v & 1);
b [1] = 0 != (v & 2);
b [2] = 0 != (v & 4);
b [3] = 0 != (v & 8);
Here's a very simple way to do it;
int main()
{
int s=7,l=1;
vector <bool> v;
v.clear();
while (l <= 4)
{
v.push_back(s%2);
s /= 2;
l++;
}
for (l=(v.size()-1); l >= 0; l--)
{
cout<<v[l]<<" ";
}
return 0;
}
Using std::bitset
int value = 123;
std::bitset<sizeof(int)> bits(value);
std::cout <<bits.to_string();
#prateek thank you for your help. I rewrote the function with comments for use in a program. Increase 8 for more bits (up to 32 for an integer).
std::vector <bool> bits_from_int (int integer) // discern which bits of PLC codes are true
{
std::vector <bool> bool_bits;
// continously divide the integer by 2, if there is no remainder, the bit is 1, else it's 0
for (int i = 0; i < 8; i++)
{
bool_bits.push_back (integer%2); // remainder of dividing by 2
integer /= 2; // integer equals itself divided by 2
}
return bool_bits;
}
#include <stdio.h>
int main(void)
{
int number = 7; /* signed */
int vbool[8 * sizeof(int)];
int i;
for (i = 0; i < 8 * sizeof(int); i++)
{
vbool[i] = number<<i < 0;
printf("%d", vbool[i]);
}
return 0;
}
If you don't want any loops, you'll have to write it out:
#include <stdio.h>
#include <stdbool.h>
int main(void)
{
int num = 7;
#if 0
bool arr[4] = { (num&1) ?true: false, (num&2) ?true: false, (num&4) ?true: false, (num&8) ?true: false };
#else
#define BTB(v,i) ((v) & (1u << (i))) ? true : false
bool arr[4] = { BTB(num,0), BTB(num,1), BTB(num,2), BTB(num,3)};
#undef BTB
#endif
printf("%d %d %d %d\n", arr[3], arr[2], arr[1], arr[0]);
return 0;
}
As demonstrated here, this also works in an initializer.