What's the difference between using static and dynamic arrays

[looopass]

Hi, I want to use 2D array in a dynamic-memory-allocation way, so I define the following in C

int row=3,column=2, i, j;
double A[3][2]={ {1,2}, {3,4}, {5,6} };
double** AD=(double **)malloc(sizeof(double *)*row);
double *ptr=(double *)malloc(sizeof(double)*row*column);

for(i=0;i<row;i++)
{
AD[i]=ptr+i*column;
for(j=0;j<column;j++)
AD[i][j]=A[i][j];
}

Then, I think, AD and A have identical element values, and the same memory address layout. But when I call a function from an IBM library, using A as the argument will give me correct result, but using AD as the argument will give wrong result. Could anyone tell me about the difference of A and AD? I mean, in which situation do I have to use the first one?

Thanks very much in advance.

Alex

[JosAH]

A picture tells more than a thousand words. Your A and AD structures are by
no means identical:

 
A: +---+---+          AD: +---+        +---+---+---+---+---+---+  
   |   |   |              | --|------> |   |   |   |   |   |   |
   +---+---+              +---+        +---+---+---+---+---+---+
   |   |   |              | --|------------------^       ^
   +---+---+              +---+                          |
   |   |   |              | --|--------------------------+
   +---+---+              +---+
 

The matrix on the left can be seen as three rows; each row has two elements
(doubles in your case). Here is the definition of a single row:

typedef double (row_t)[2];
 

So if you allocate three rows:

row_t* ad= calloc(sizeof(row_t), 3);
 

your ad structure is identical to your (statically allocated) a stucture now; it now
points to three rows adjacent in memory.

kind regards,

Jos

[looopass]

thanks for you reply.

Your picture did a great job. Instead of passing AD to the function, I should pass (AD)[0] to it. And now it works. Thanks

[JosAH]

thanks for you reply.

Your picture did a great job. Instead of passing AD to the function, I should pass (AD)[0] to it. And now it works. Thanks

I was just about to reply to your previous incarnation of your reply but I noticed
you edited it. For a two dimensional arrray which consists of a single block of
memory either build a 'dope' vector like you did and pass the very first row element
around, i.e. you have made the row pointers explicit pointers or cast a bit here
and there and define a row type like I did. My trick doesn't need a 'dope' vector,
your trick doesn't need the new type.

kind regards,

Jos

[looopass]

Thanks a lot Jos. you solved my problem