Marcus Marcus - 1 year ago 86
C Question

Counting primes up to 18 digits optimization

I have a task at school to count number of primes up to 10^18 in under 2 minutes and with no more than 2 GB of memory to use. For the first try I've implemented a segmented sieve with the following optimizations:


  • used 32 bit integers to store the primes in a segmented as bits

  • didn't store the odd numbers

  • segmented the sieve into parts of size sqrt(n)

  • counting the composite primes when marking them so I don't have to loop the sieve again

  • used dynamic memory allocation to store the first primes up to sqrt(n) (in this case I've created a Queue in C to store the primes)



The problem is that for counting the primes up to 10^9 on my computer(which has pretty decent specs) it takes 13 seconds, therefore for 10^18 would take me some days.

My question is, is there some optimization that I am missing, or is there a better and faster way to count primes up to a number? The code:

#include <stdio.h>
#include <stdlib.h>
#include <math.h>
#include <time.h>
#include <string.h>

typedef signed char int8_t;
typedef signed short int int16_t;
typedef signed int int32_t;
typedef signed long long int int64_t;

typedef unsigned char uint8_t;
typedef unsigned short int uint16_t;
typedef unsigned int uint32_t;
typedef unsigned long long int uint64_t;

#define SIZE 32
#define DEBUG

#define KRED "\x1B[31m"
#define KGRN "\x1B[32m"
#define KYEL "\x1B[33m"
#define KBLU "\x1B[34m"
#define KMAG "\x1B[35m"
#define KCYN "\x1B[36m"
#define KWHT "\x1B[37m"
#define RESET "\033[0m"

struct node {
uint64_t data;
struct node* next;
};

struct queue {
struct node* first;
struct node* last;
uint32_t size;
};

typedef struct node Node;
typedef struct queue Queue;

/* Queue model */
uint8_t enqueue(Queue* queue, int64_t value) {
Node* node = (Node*)malloc(sizeof(Node*));

if (node == NULL)
return 0;

node->data = value;
if (queue->last)
queue->last->next = node;

queue->last = node;
if (queue->first == NULL)
queue->first = queue->last;

queue->size++;
return 1;
}

uint64_t dequeue(Queue* queue) {
Node* node = queue->first;
uint64_t save_data = node->data;

if (queue->size == 0)
return 0;

queue->first = queue->first->next;
queue->size--;
free(node);

return save_data;
}

Node* queue_peek(Queue* queue) {
return queue->first;
}

uint32_t queue_size(Queue* queue) {
return queue->size;
}

Queue* init_queue() {
Queue* queue = (Queue*)malloc(sizeof(Queue*));

queue->first = queue->last = NULL;
queue->size = 0;

return queue;
}

/* Working with bit arrays functions */
uint8_t count_set_bits(uint64_t nbr) {
uint8_t count = 0;

while (nbr) {
count++;
nbr &= (nbr - 1);
}

return count;
}

uint8_t get_bit(uint32_t array[], uint32_t position) {
const uint64_t mask = 1U << (position % SIZE);

return array[position / SIZE] & mask ? 1 : 0;
}

void clear_bit(uint32_t array[], uint32_t position) {
const uint64_t mask = ~(1U << (position % SIZE));

array[position / SIZE] &= mask;
}

void set_bit(uint32_t array[], uint32_t position) {
array[position / SIZE] |= (1U << (position % SIZE));
}

/* Solve the problem */
Queue* initial_sieve(uint64_t limit) {
Queue* queue = init_queue();
uint64_t _sqrt = (uint64_t)sqrt(limit);
uint32_t *primes = (uint32_t*)calloc(_sqrt / SIZE + 1, sizeof(uint32_t));

set_bit(primes, 0);
// working with reversed logic, otherwise primes should all me initialized to max uiint64_t

enqueue(queue, 2);
for (uint64_t number = 3; number <= _sqrt; number += 2) {
if (get_bit(primes, number / 2) == 0) {
enqueue(queue, number);

for (uint64_t position = number * number; position <= _sqrt; position += (number * 2)) {
set_bit(primes, position / 2);
}
}
else
set_bit(primes, number / 2);
}

return queue;
}

uint64_t count_primes(uint64_t limit) {
uint64_t start, end, delta;
uint64_t non_primes_counter, initial_size;
uint32_t *current_sieve;
Queue* queue;

queue = initial_sieve(limit);
initial_size = queue->size;
start = delta = (uint64_t)sqrt(limit);
end = 2 * start;
non_primes_counter = 0;

printf("Limits: %llu -> %llu\n", start, end);
while (start < limit) {
Node* prime = queue->first->next; // pass 2 since only odd maps are represented in the sieve
uint64_t count = 0;

current_sieve = (uint32_t*)calloc(delta / SIZE + 1, sizeof(uint32_t));
// memset(current_sieve, 0, sizeof(uint32_t) * delta);

while (prime != NULL) {
uint64_t first_composite = start / prime->data * prime->data;

// calculate the first multiple of the given prime in the interval
if (first_composite < start)
first_composite += prime->data;
if ((first_composite & 1) == 0)
first_composite += prime->data;

// set all the composites of the current prime in the given interval
for (uint64_t number = first_composite; number <= end; number += (prime->data) * 2) {
const uint64_t position = (number - start) / 2;

if (get_bit(current_sieve, position) == 0) {
set_bit(current_sieve, position);
count++;
}
}

// free(current_sieve);
prime = prime->next;
}

non_primes_counter += count;
start += delta;
end += delta;

if (end > limit)
end = limit;
}

uint64_t total = (limit - delta + 1) / 2 - non_primes_counter;

printf("%sTotal composites and initial size: %llu %llu %s\n", KCYN, non_primes_counter, initial_size, RESET);
printf("%sTotal primes: %llu %s\n", KCYN, total, RESET);
return queue->size + (limit - delta + 1) / 2 - non_primes_counter;
}

/* Main */
int main(int argc, char **argv) {
clock_t begin, end;
double time;

if (argc < 2) {
printf("Invalid number of parameters\n");
printf("Program will exit now.\n");
return 0;
}

begin = clock();
printf("%sNumber of primes found up to %s%s: %s%llu.\n%s", KWHT, KCYN, argv[1], KYEL, count_primes(atoll(argv[1])), RESET);
end = clock();
time = (double)(end - begin) / CLOCKS_PER_SEC;

printf("%sTotal time elapsed since the starting of the program: %s%lf seconds.\n%s", KWHT, KYEL, time, RESET);
return 0;
}


Thanks, Marcus

Answer Source

You need to count the number of primes, not to find them all (there are too many of them). This called a Prime-counting function.

In mathematics, the prime-counting function is the function counting the number of prime numbers less than or equal to some real number x. It is denoted by π(x).

There are plenty of methods that calculate this function. Look at this Wolfram page with methods comparison. It seems it would be a hard task to accomplish this in two minutes.

As was mentioned in the comments there is also a great answer at math.stackexchange, I think it would be helpful.

Recommended from our users: Dynamic Network Monitoring from WhatsUp Gold from IPSwitch. Free Download