Algorithm Optimization (Prime Factorization)

Question

Before starting let me say: It's not homework, just plain, old, fun.

Now, I'm trying to come up with an algorithm that can answer this question 1/x + 1/y = 1/n!.

And as you can see by the link above, the author asked only for hints and not the actual answer, so I would kindly ask for the same.

I simplified the expression until (x - n!)(y - n!) = (n!)^2 as suggested by one of the answers, and by that time I understood that the number of combinations of (x,y) pairs is the same as the number of divisors of n!^2 (correct me if I'm wrong here).

So, as suggested by the accepted answer, I'm trying to get the multiplication of all the factors of each prime composing N!^2.

I've come up with some code in C using trial division to factorize N!^2 and the Sieve of Eratosthenes to get all the prime numbers up to sqrt(N!^2).

The problem now is memory, I have tried with N = 15 and my Mac (Quad Core 6GB of memory) almost died on me. The problem was memory. So I added some printf's and tried with N=11:

Sieve of Eratosthenes took 13339.910000 ms and used 152 mb of memory
n= 11; n!^2 = 1593350922240000; d = 6885
[2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,3,3,3,3,3,3,3,3,5,5,5,5,7,7,11,11]

The list is all the prime factors of N!^2 (besides 1 and N!^2 of course).

I would like some hints on how to minimize memory consumption and possible optimizations.

Code bellow, it was just a quick experiment so I'm sure it can be optimized.

#include <stdio.h>
#include <stdlib.h>
#include <math.h>
#include <strings.h>
#include <sys/time.h>
#include <assert.h>

//Linked List
struct node {
    struct node * next;
    long val;
};

void addValue(struct node *list, long val) {
    struct node *n = list;

    if (n->val == -1) {
        n->val = val;
        return;
    }

    while (n->next) {
        n = n->next;
    }

    struct node *newNode = malloc(sizeof(struct node));
    newNode->val = val;
    newNode->next = NULL;
    n->next = newNode;
}

void freeLinkedList(struct node *list) {
    struct node *c = list;
    if (!c) return;
    struct node *n = c->next;
    free(c);
    freeLinkedList(n);
}

void printList(struct node *list) {
    struct node *n = list;
    printf("[");
    while (n) {
        printf("%ld", n->val);
        n = n->next;
        if (n) {
            printf(",");
        }
    }
    printf("]\n");
}
//-----------


int fac(int n) {
    if (n == 1) return 1;
    return fac(n-1)*n;
}

//Sieve of Eratosthenes
int sieve_primes(long limit, long **list) {
    struct timeval t1;
    struct timeval t2;
    double elapsedTime = 0;
    gettimeofday(&t1, NULL);

    assert(limit > 0);

    //Create a list of consecutive integers from 2 to n: (2, 3, 4, ..., n).
    long arrSize = limit-1;
    long *arr = malloc(sizeof(long)*arrSize);

    long c = 2;
    for (long i = 0; i < arrSize; i++) {
        arr[i] = c++;
    }   
    assert(arr[arrSize-1] == limit);


    for (long i = 0; i < arrSize; i++) {
        //Let p be equal to the first number not crossed
        long p = arr[i];    
        if (p == 0) continue;

        //Starting from p, count up in increments of p and mark each of these numbers greater than p itself in the list. 
        for (long f = p+p; f < arrSize; f+=p) {
            arr[f] = 0;
        }       
    }

    *list = arr;


    gettimeofday(&t2, NULL);

    elapsedTime = (t2.tv_sec - t1.tv_sec) * 1000.0;      // sec to ms
    elapsedTime += (t2.tv_usec - t1.tv_usec) / 1000.0;   // us to ms
    printf("Sieve of Eratosthenes took %f ms and used %lu mb of memory\n",elapsedTime, (arrSize * sizeof(int))/1024/1024);
    return arrSize;
}

void trial_division(struct node* list, long n) {    if (n == 1) {
        addValue(list, 1);
        return;
    }
    long *primes;
    long primesSize = sieve_primes(sqrt(n), &primes);   

    struct timeval t1;  
    struct timeval t2;
    double elapsedTime = 0;
    gettimeofday(&t1, NULL);
    for (long i = 0; i < primesSize; i++) {
        long p = primes[i];
        if (p == 0) continue;
        if (p*p > n) break;
        while (n % p == 0) {
            addValue(list, p);
            n/=p;
        }       
    }
    if (n > 1) {
        addValue(list, n);
    }
    free(primes);
}

int main(int argc, char *argv[]) {
    struct node *linkedList = malloc(sizeof(struct node));
    linkedList->val = -1;
    linkedList->next = NULL;


    long n = 11;
    long nF = fac(n);
    long nF2 = nF*nF;
    trial_division(linkedList, nF2);            

    long multOfAllPrimeFactors = 1;
    struct node *c = linkedList;
    while (c) {
        long sumOfVal = 2;
        long val = c->val;              
        c = c->next;
        while(c) {
            long val2 = c->val;
            if (val == val2) {
                sumOfVal++;
                c = c->next;
            } else break;           
        }
        multOfAllPrimeFactors*=sumOfVal;
    }       

    printf("n= %ld; n!^2 = %ld; d = %ld\n", n,nF2, multOfAllPrimeFactors);
    printList(linkedList);  

    freeLinkedList(linkedList);

}

EDIT:

As an example I will show you the calculation for getting all the possible positive integer solutions to the initial equation:

3!^2 = 36 = (3^2*2^2*1^0)

So there are (1+2)(1+2)(1+0)=9 possible positive integer solutions to the diophantine equation. Double if you count negative integers. I'm using WolframAlpha to be sure.

EDIT 2:

I think I just found out "what a factorial is", I'm getting this very interesting output:

3! = [2,3]
3!^2 = [2,2,3,3]
3!^3 = [2,2,2,3,3,3]
3!^4 = [2,2,2,2,3,3,3,3]

Thanks :D

Answer 1

The trick here is to recognize exactly what a factorial N! is. It's a product of all the numbers from 1 to N. Which is already a huge step forward.

So what you need to do, is just to prime factorize each of the numbers from 1 to N.

In this sense, you don't need to sieve up to N!. Instead, just sieve up to sqrt(N). And the rest is just merging all your prime factors.