For a prototype of the form
  ReadMatrix(float **A, int n, file input-file);         //Definition-Old

you need to return as int the dimension i.e.

int  ReadMatrix(float **A, int n, file input-file) {

   //do whatever

   return(dimension-read-from-input-file);
}
 i.e. dimension n is not used.

  If you want to change it and store it in the parameters
you need to change          V------attentionhere
  ReadMatrix(float **A, int *n, file input-file);        //Definition-New

In that case you can have ReadMatrix as void.
void  ReadMatrix(float **A, int n, file input-file) {

   //do whatever

   *n =dimension-read-from-input-file;
}

 
For the other **A
 yes you
 define a
  float *A;
 and call
  ReadMatrix(&A, &n,input-file) //Definition-New

 However A is a pointer to a pointer to a float.
 The following definition is WRONG
  A = (float**)malloc(sizeof(float)*nMatrixSize*nMatrixSize);
  in the sense that the block allocated has an address (first byte)
  available through
     (float*)malloc(sizeof(float)*nMatrixSize*nMatrixSize);
  but that address is not stored in a variable whose address could then be
  accessed through by a  & .

 The correct way is to do

  ReadMatrix(float **A, int *n ... ) {

{
   float * MyA;

   MyA = (float *)  blah blah;

   *A = myA; //returns in *A the address of the block allocated.
 }


-----
 HW 3 (only Programming 1) has been updated to reflect to Definition-New definition.
 Java users won't be affected, and C/C++ users can still use the old definition if so
 desire (by returning the dimension as return date).


alex -- Tue 17 Oct, 2:29pm