Integer types and overflow

[nullcoding]

Hello!

I'm very much a beginner at programming. I'm teaching myself by trial-and-error whilst my partner is taking a basic C++ class at Uni. He wrote most of this program, but I've done a fair bit of the nitpicky stuff, like stdout to text file and the initial looping function. I'm pretty familiar with basic syntax, but I know about as much as someone would after a week of class. Suffice it to say, I'm very new and inexperienced!

So this program we've got is a basic Proth number tester. It takes the formula N=k*(2^n)-1 and asks user for k as well as lower and upper bound of n. It then tests N for k at all values of n between n(lower) and n(upper). It then prints those to a neat text file.

The main problem that irks me at the moment is that I don't know how to overcome an overflow error. The best way I can describe it is that it's similar to what I've seen in Java - cannot have more than 2^31 of "something."

Hence, our results file might look like this:

You chose 5 as your value of k.
You chose 28 as your lower bound and 29 as your upper bound.

Here are the results...

5*2^28-1 is 1342177279.

1342177279 is divisible by 17989 and has the factors 17989 and 74611.

1342177279 appears to be composite.


5*2^29-1 is -2147483648.

-2147483648 is divisible by 4 and has the factors 4 and -536870912.

-2147483648 appears to be composite.


All done.

I have heard that we ought to be using unsigned long. I've seen some examples of this, but can't input it in a syntactically correct manner.

From my basic understanding, the length of the integer N=5*2^29-1 is more than 10 digits and therefore cannot be represented by our program.

How do I make the program able to deal with larger (or very large) numbers?

Also, is there a way to exceed 2^64 (which I understand to be the limit of unsigned long in C++) without being on a 64-bit system?

Thanks in advance for any help/explanation. I recognize this is quite the newbie issue, but I can't figure out proper implementation no matter how many articles I read.

[zb0S216C]

There's two possible scenarios when an overflow can occur (as far as I know of):

1) When an integer is assigned a value that's beyond its maximum value
2) When the program pushes too much information onto the stack which exceeds the 1MB limit.

Without your program, it's hard to determine.

_Wazzak

[nullcoding]

Luckily, I made a Google Code project so we can work on it together across distances (he's 100 miles away). The code's in the source section, here: http://code.google.com/p/gjsieve/source/browse/main_4.cpp

[zb0S216C]

I can't see any stack violations, so it must be a numerical range violation. There's two solutions I can think of at this time:

1) Impose integral constraints on the input values
2) Use the long long type for lower, k and upper. This gives roughly twice the range of signed int.

_Wazzak

[LB]

Tip: You always use unsigned types unless you have a need for negative numbers.

[nullcoding]

The second option sounds like a lot of things I've read about.

The only issue is that I can't figure out where to put that. Currently what I have is inline float pow(float x, float y) {return pow((long double)x,(long double)y);} and I understand I can't simply replace float with long long.

Does this simply involve changing int in int lower, k, upper; to long long? Because that still gives overflow errors. I suppose I'm missing something.