C++ segmentation fault

Asuralm · 12-05-2007, 10:39 AM

Hi all:

When I try to declare a large matrix type or vector type, say a matrix of 3000*3, or a vector 3000*1 something like that, when running the program it reported the segmentation fault. When I try a small matrix like 10*3 then it's fine.

Does anyone know why it is please? Is it because of my previlidge?

The matrix and vector type I used are from the numerical recipes.

Any help would be appreciated!!!

Thanks in advance

Hivemind · 12-05-2007, 10:47 AM

You didn't provide much detail so I can only guess but one cause may be that you're trying to store this huge matrix on the stack, which is not big enough to hold it. Allocate it on the heap instead using new.

sydney-troz · 12-05-2007, 11:17 AM

Can you post the relevant code?

Asuralm · 12-05-2007, 12:42 PM

Yeah, here is the code:

Code:

template <class T>
class NRVec{
private:
    int nn;     // size of array. upper index is nn-1
    T *v;
public:
    NRVec();
    explicit NRVec(int n);      // zero-based array
    NRVec(const T &a, int n);   // initialize to constant value
    NRVec(const T *a, int n);   // initialize to array
    NRVec(const NRVec &rhs);   // copy constructor
    NRVec & operator=(const NRVec &rhs);    //assignment
    NRVec & operator=(const T &a);      // assign a to every element
    inline T & operator[](const int i);     // i'th element
    inline const T & operator[](const int i) const;
    inline int size()const;
    ~NRVec();
};

And I have built them into a library.
In another project, I invoke the NRVec class like this.

Code:

typedef NRVec<float> Vec_SP;
.
.
.
In some function, I declare an array lie

Vec_SP myarray(3000);

Then it failed. It only work if I declare a very small array like
Vec_SP myarray(10);

Asuralm · 12-05-2007, 12:48 PM

Code:

template<class T>
NRVec<T>::NRVec() : nn(0), v(0) {}

template<class T>
NRVec<T>::NRVec(int n) : nn(n), v(new T[n]){}

template <class T>
NRVec<T>::NRVec(const T& a, int n) : nn(n), v(new T[n])
{
    for(int i =0; i < n; i++)
        v[i] = a;
}

template <class T>
NRVec<T>::NRVec(const T *a, int n) : nn(n), v(new T[n])
{
    for(int i =0; i<n; i++)
        v[i] = *a++;
}

template <class T>
NRVec<T>::NRVec(const NRVec<T> &rhs) : nn(rhs.nn), v(new T[nn])
{
    for(int i =0; i<nn; i++)
        v[i] = rhs[i];
}

Here is the definition of the NRVec class. I got this from the numerical recipes.

sydney-troz · 12-05-2007, 05:09 PM

Hmm, it compiles and runs fine on my amd64 (runs fine when compiled as 64bit and 32bit (with -m32 flag)). The code looks good, have you tried it on another system? If so, did it work? If not, what compiler are you using? If so, what are your specs?

Also, could you email me the entire source file? I want to try compiling your exact code. s_troz <at> yahoo <dot> ca

Thanks