stack overflow error in C++ program

问题:

So i have this complex class , and i want to have an 2d array of complex numbers this is part of the code not all the code

class Complex {
public:
    /* construction/destruction */
    Complex(double r, double i)     { this->r = r; this->i = i; }
    Complex()                       { r=0.0; i=0.0; }
    ~Complex()                      { r=0.0; i=0.0; }
        /* operations */
    Complex operator+(Complex &c)   { return Complex( r+c.r, i+c.i ); }
        double r, i;
};

int main()
{
const int HEIGHT = 256;
const int WIDTH = 256;
Complex G[HEIGHT][WIDTH];
}

so the line Complex G[HEIGHT][WIDTH]; is the line that causes the problem , any idea why ?

回答1:

Visual studio defaults to 1MB stack size, it looks like:

Complex G[HEIGHT][WIDTH];

will be just about 1MB, you can modify this using /F and the document says (emphasis mine):

Without this option the stack size defaults to 1 MB. The number argument can be in decimal or C-language notation. The argument can range from 1 to the maximum stack size accepted by the linker. The linker rounds up the specified value to the nearest 4 bytes. The space between /F and numberis optional.

The most obvious alternative would be to use dynamic memory allocation via new or std::vector.

Visual Studio as far as I know actually has one of the smaller default stack sizes:

platform    default size       
=====================================
SunOS/Solaris  8192K bytes
Linux          8192K bytes
Windows        1024K bytes
cygwin         2048K bytes
Mac OS X       8192K bytes

回答2:

Any idea why?

Some compilers default the stack size to 1MB. You are allocating 65536 Complex objects that occupies 2 * sizeof(double) memory each. Assuming double to be 8 bytes (this information is implementation defined) you are effectively trying to allocate 16 * 65536 bytes (without considering possible paddings), which are 1048576 bytes, that causes the overflow.

An alternative is using dynamic allocation with a wrapper, that simulates a bi-dimensional array indexing, along the lines of this one:

template<std::size_t A, std::size_t B>
class G {
private:
    std::unique_ptr<Complex[]> mem;
public:
    G() : mem(new Complex[A * B]) {}
    Complex& operator()(std::size_t a, std::size_t b) {
        return mem[a * B + b];
    }
    Complex  operator()(std::size_t a, std::size_t b) const {
        return mem[a * B + b];
    }
};

Then you program simply becomes:

int main(int, char*[]) {
    G<256, 256> g;
    g(0, 0) = ...;
}

Of course you can generalize your wrapper G for a generic type with templates, but that's outside the scope of this answer.

---

On a side note, you destructor:

~Complex() { r=0.0; i=0.0; }

is useless. Don't re-initialize memory that will be destroyed anyway when it leaves the scope.