exp: OVERFLOW error

**spiit231** · 03-04-2008

Hi all,
I am using Borland C++ 4.5 on 32-bit windows machine. I am getting "exp: OVERFLOW error" for my code. can anybody tell me how to deal with it.

I am not posting my

Code:

struct MATRIX {
	int rows;
	int cols;
	double **t;
};

...
...

double norm(struct MATRIX t, double s){...} 
\\returns s-norm

double scalar(struct MATRIX t){...} 
\\returns the only element for 1X1 matrix

struct MATRIX ind_response(struct MATRIX t, int index){...}
\\returns a structure containing a column matrix having t.cols number of rows

struct MATRIX multifun(struct MATRIX a, struct MATRIX b, char ch)
\\return multiplied value of a.t and b.t matrices when ch == '*'

struct MATRIX transpose(struct MATRIX t)
\\returns transpose of matrix

double fun(struct MATRIX x, struct MATRIX y, struct MATRIX z, int l){
	double fval, sum = 0;
	float bx, ybx;
	int i;

	for(i = 0; i < y.rows; i++){
		bx = scalar(multifun(transpose(z), ind_response(x, i), '*'));
		ybx = *(*(y.t+i)+0)* bx;
		sum = sum - (ybx - log(1 + exp(bx)));
		}
	fval = sum + l* norm(z, 2.0);
	return fval;
	}
...

main(){
...
}

**Elysia** · 03-04-2008

Main returns int. Plus \\ is not a comment. // is. Although in C, it's /* and */.
And you failed to mention where the problem is?
But overflow means you need a bigger data type, that's all. That or you need to stop doing ridiculously high calculations so the number won't overflow.

**spiit231** · 03-04-2008

Thanks for the suggestion.

Sorry for my silly mistakes. I have given the comments for describing the functions for this thread.
Basically I dont know where the problem is. Borland displays a pop-up error message. I am showing this part of the code because only here i use math functions.
the data type i used is double. should i use long double?

**Brad0407** · 03-04-2008

If you code overflows, it made a number that was too big for it to store in whatever datatype you were trying to store the number in. It's possible that you're doing "ridiculously high calculations" (Obviously factorials would fall into this category), but since doubles numbers up to something ridiculously high like 10^300 or something, I think there could be other problems. Some things that come to mind are that you made a typo on one of the calculations and it's doing unexpected things. Also, if you don't initialize the number, it could start out really large. If you're sure that you aren't making any errors, you can use a bigger type. long double could be a solution but of course that's not guarenteed because it depends on the compiler and target machine. And you could always make your own way of handling floating point numbers.
[edit]Post your entire code. If it's really big, attach it[/edit]

**matsp** · 03-04-2008

What is the value of "bx" when it fails? That is presumably the cause of the problem. Using "long double" if the compiler supports that may solve the problem, but if you have a large enough value, any floating point type will overflow.

--
Mats

**spiit231** · 03-05-2008

I think this is due to exp() function which is unable to handle large float value. Consider the code below:

Code:

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

#define z 8113.48

main(){
	long double x;
	x = exp(z);
	printf("%Lf", x);
	return 0;
	}

this also displays same overflow error. Now please suggest me how to handle large float value in this case.

**Elysia** · 03-05-2008

For the love of... I thought I taught you to use int main. Please use int main!
Long double is the largest data type C can handle. x is initialized, as well.
%Lf is not used to print doubles either. %f is.
And if you get overflow, it means you're doing too high calculations, simply put.

**spiit231** · 03-05-2008

Is it correct?

Code:

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

#define z 8113.48

int main(){
	long double x;
	x = exp(z);
	printf("%f", x);
	return 0;
	}

But %f is not working !!!!!
Also my problem is unsolved.

**Elysia** · 03-05-2008

%f works. You're getting overflow, and there's not much to do about that unless you magically invent another type that holds larger numbers.

**matsp** · 03-05-2008

According to Windows Calculator 2.72 to the power of 8113 is 4.5e+3525, which is WAY outside of the range for double (by about ten orders of magnitude). If you can find a "exp()" function that takes a long double, you may be able to get it to work. According to "google man exp", the function you want could be called "expl()" - but that may depend on the compiler library, so don't hold me responsible if that doesn't work.

--
Mats

**matsp** · 03-05-2008

Elysia: Note that the value is "long double", which does indeed need to be prefixed to %lf or some such. Which is correct depends on the compiler version (I think C99 standardizes it).

--
Mats

**Elysia** · 03-05-2008

News to me, but anyway, as mentioned, the problem is overflow.

**matsp** · 03-05-2008

Originally Posted by Elysia

News to me, but anyway, as mentioned, the problem is overflow.

Yes, having the wrong format for printf() will only change how printf behaves, not how the floating point unit and the exp() function behaves, that is correct.

--
Mats

**spiit231** · 03-05-2008

expl() also not working. Suppose i want to create a new data type that can handle large floating values. Since i have no idea about this area...can anybody lighten me up?

Another thing, what is the range of x in exp(x) for double and long double data types. I'm using Borland C++ 4.5.

**Elysia** · 03-05-2008

I can suggest you don't use Borland, if you can.
See cpwiki.sf.net/Integrated_Development_Environment for a list of other IDEs.

Thread: exp: OVERFLOW error

Thread Tools

Search Thread

Display

exp: OVERFLOW error

Similar Threads

Compiling sample DarkGDK Program

Testing some code, lots of errors...

Another syntax error

Post...

Stupid compiler errors