Thread: Prime #'s question. "Floating point exception" error.

Code:

#include<stdio.h>
#include<math.h>
long long int testNum;
long long int testAgainst;
short int isNumPrime;

int isPrime(long long int testNum, long long int testAgainst);

int main()
{
//9223372036854775807 = largest int
for(testNum=3;testNum<9223372036854775807;testNum++)
        {
        if(isPrime(testNum,testAgainst)==1)
                {
                printf("%lli\n",testNum);
                }
        }
return 0;
}

int isPrime(long long int testNum,long long int testAgainst)
{
int isNumPrime;
for(isNumPrime=1;testAgainst!=sqrt(testNum);testAgainst++)
        {
        if(testNum%testAgainst==0)
                {
                isNumPrime=1;
                } else {
                isNumPrime=0;
                break;
                }
        }
return isNumPrime;
}

I wrote a program to count from 3 to the highest long long int my compiler can handle. Only problem is, I'm not even sure if it works as I get a 'floating point exception' error on compile.

I'm guessing it's for one of three reasons, but I can't point it out, maybe someone can help me there?

Trap representations of signed types
Overflow on signed types
Division by zero

If I divided by zero, I would be so happy.

Forcing a compiler to divide by zero is not exactly a cause for celebration.

The magic value (9223372036854775807) will be treated as type int by default. Change it to 9223372036854775807LL. Or, better yet, #include <limits.h> and replace that magic value with LLONG_MAX. That will work with any C compiler that supports long long types.

A more fundamental problem in your code is that you are assuming a long long type can be converted to floating point without loss of information .... which is pretty unlikely for any compiler that supports an 8-byte double and 8-byte long integral types (i.e. a significant proportion of real-world compilers). sqrt() is for floating point, and is not guaranteed to work particularly accurately for integral types - particularly large integral values. Have a look at this post for a hint about an alternative that would work better....

Unless you want to work with negative values, it might pay to use long long unsigned variables.

Note that I haven't even bothered to check whether your test for primality is valid. But you will need to check that....

Thank you for taking a peek grumpy.

Code:

#include<stdio.h>
#include<math.h>
#include<limits.h>

long long int testNum;
long long int testAgainst;
short int isNumPrime;

int isPrime(long long int testNum, long long int testAgainst);

int main()
{
for(testNum=3;testNum<LLONG_MAX;testNum++)
        {
        if(isPrime(testNum,testAgainst)==1)
                {
                printf("%lli\n",testNum);
                }
        }
return 0;
}
int isNumPrime;
for(isNumPrime=1;testAgainst!=testNum-1;testAgainst++)
        {
        if(testNum%testAgainst==0)
                {
                isNumPrime=1;
                } else {
                isNumPrime=0;
                break;
                }
        }
return isNumPrime;
}

Changed it, but I'm still getting the same error.

Edit: I'm not happy that I may have forced my compiler to divide by zero, rather I would be happy that I've possibly done something I've never done before.

Edit again: I found the particular error, now I've got a new one. I'll reply on this thread if I can't get past it. The problem was I initialized int testAgainst, but I never set a value above zero, so I was, in fact, forcing my compiler to divide by zero after all lol.

Updated the source and now it hangs after printing two numbers.

Code:

#include<stdio.h>
#include<math.h>
#include<limits.h>
long long int testNum;
long long int testAgainst=2;
short int isNumPrime;

int isPrime(long long int testNum, long long int testAgainst);

int main()
{
//9223372036854775807 = largest int
for(testNum=3;testNum<LLONG_MAX;testNum++)
	{
	if(isPrime(testNum,testAgainst)!=0)
		{
		printf("%lli\n",testNum);
		}
	}
return 0;
}

int isPrime(long long int testNum,long long int testAgainst)
{
for(isNumPrime=1;testAgainst!=testNum-1;testAgainst++)
	{
	if(testNum%testAgainst!=0)
		{
		isNumPrime=0;
		break;
		}
	}
return isNumPrime;
}

Output:

Code:

andrew@home:~/Documents/c/isPrime$ ./test
3
4

You have confused yourself by having arguments of isPrime() with the same name as global variables. So isPrime() does not affect the global variables.

Originally Posted by grumpy

You have confused yourself by having arguments of isPrime() with the same name as global variables. So isPrime() does not affect the global variables.

I'm confused by that. I moved the variables from the top and inside int main() and int isPrime() and it's still happening just as before. Or were you referring to something else?

Code:

//snip
#include<limits.h>
int isPrime(long long int testNum,long long int testAgainst);
//snip

Code:

int main()
{
long long int testAgainst = 0;
long long int testNum = 0;
//snip

Code:

int isPrime()
{
short int isPrime=1;
//snip

Started over, rewrote the majority of it using a dry-erase board. It works rather well now. Any thoughts, comments, suggestions, etc. are welcome and appreciated.

Code:

#include<stdio.h>
#include<limits.h>

int main()
{
short int yesno;
long long int i,j;
long long int x=3;
long long int y=LLONG_MAX;

FILE *fp;
fp = fopen("output","w+");
fprintf(fp,"1, 2, ");

for(i=x;i<=y;i++)
{
yesno=0;
for(j=2;j<=i/2;j+=2)
	{
	if(i%j==0)
		{
		yesno=1;
		break;
		}
	}
	if (yesno==0)
		{
		fprintf(fp,"%lld, ",i);
		}
	}
fclose(fp);
return 0;
}

Looks like you're testing every number for all even factors. That's going to include 9 in the output of primes, and a whole lot more that only have odd factors.
Also, you're testing every number for primeness, whereas you may as well only test odd numbers above 2 and skip every even number.
Can you see what you did wrong there now?

Why don't you only go up to the square root of the number being tested like you did before? Just don't use the srqt function. Instead change the loop condition to j*j<=i and it'll be a fair bit faster. You would have found that earlier if you read to the end of the thread that grumpy linked to.

I did actually, I just didn't update this since. I've made a few improvements to the code, but yeah. One of them is only checking odd numbers, then I implemented the sqrt without sqrt. It's been running for about 20 minutes and has a 8mb text file :P

In the above post I said it now takes me 20 minutes to make a file of 8 gigs. Not so. That was before I implemented the changes. Now it takes about 30 seconds

My parents think I'm hacking... it's kind of embarrassing really. They see a terminal window with a bunch of numbers flying by and they automatically assume the worst. Anyway, here's the updated source so you can see what I've changed.

Code:

#include<stdio.h>
#include<limits.h>

int main()
{
short int yesno;
long long int i,j;
long long int x=3;
long long int y=LLONG_MAX;

FILE *fp;
fp = fopen("output","w+");

printf("1, 2, ");
fprintf(fp,"1, 2, ");

for(i=x;i<=y;i++)
{
yesno=0;
for(j=2;j*j<=i/2;j+=2)
	{
	if(i%j==0)
		{
		yesno=1;
		break;
		}
	}
	if (yesno==0)
		{
		printf("%lld\n",i);
		fprintf(fp,"%lld, ",i);
		}
	}
fclose(fp);
return 0;
}

You didn't pay attention to the first part of my post.
Check the contents of the file and you'll see that 9, 15, 21 etc are listed amoung the primes, because you still have it very wrong.
Read my post again carefully, and if you still can't see the problem then ask for clarification.

Yes I did, I just copy/pasted from the backup source file lol. You guys are going to have to forgive me for that. Nano isn't exactly the friendliest of text editors when it comes to mouse usage and opening a file. It opened ~test.c instead of test.c and I didn't look :P

Edit:

~ goes to your home directory, ctrl+tab (tries to) finish the command for you. Backups are prefixed with a ~ by default and it went downhill from there. Before I fixed the code I had been up for 36 hours, that's my only excuse there :P

The variable names are what I'm used to in javascript.

a, b, c, i, j, k, x, y, z are my usual variables for loops, less to type out but I suppose you're right, it could cause a bit of confusion. I'll see if I can't come up with a different set of variables.

I also said 8 gigs... meant mibs (how Ubuntu refers to megabytes)... As you all may or may not have been able to tell. I've been very tired lately.

THIS is the source code I meant to post lol. I had to open up all of them in gedit and look them over (this is when backups begin to look like a bad idea lol).

Code:

#include<stdio.h>
#include<limits.h>

int main()
{
short int yesno;
long long int i,j;
long long int x=3;
long long int y=LLONG_MAX;

FILE *fp;
fp = fopen("output","w+");

printf("1, 2, ");
fprintf(fp,"1, 2, ");

for(i=x;i<=y;i+=2)
{
yesno=0;
for(j=3;j*j<=i;j+=2)
	{
	if(i%j==0)
		{
		yesno=1;
		break;
		}
	}
	if (yesno==0)
		{
		printf("%lld, ",i);
		fprintf(fp,"%lld, ",i);
		}
	}
fclose(fp);
return 0;
}

Of course I would like to know if I've got some error in this. I've looked over all of the ones from 1 to ~8000 before I got bored with it.

Also, I am working on a way to limit the file size to ~512kb per file. I'm going to look up more information on reading file size or limiting output to files before I do anything else with it though... if I have any questions about that I'll start a new thread.

Here's some output in case anyone wants to test:

Code:

17613047
17613049
17613061
17613091
17613097
17613137
17613143
17613151
17613191
17613199
17613223
17613227
17613251
17613259
17613263
17613269
17613283
17613293
17613307
17613311
17613341
17613371
17613383
17613389
17613413
17613451
17613457
17613473
17613481
17613493
17613521
17613523
17613539
17613553