Mex File dot product - c

Question

I'm trying to implement some elementary linear algebra routines in MEX files in C for practice, and I'm stuck with dot products. Here's what I have so far:
#define char16_t UINT16_T //shenanigans with the compiler
#include "mex.h"
void dotProd(double *a, double *b, double z, mwSize n)
{
mwSize i;
for(i=0;i<n;i++){
z+=a[i] * b[i];
}
}
/* The gateway function */
void mexFunction( int nlhs, mxArray *plhs[],
int nrhs, const mxArray *prhs[])
{
double z=0; //Output scalar
double *b, *a; //Input vectors
int n;
a = mxGetPr(prhs[0]); //pointer to a
b = mxGetPr(prhs[1]); //pointer to b
n = mxGetM(prhs[0]);
// Create output
plhs[0] = mxCreateDoubleScalar(z);
dotProd(a,b,z,(mwSize)n);
}
The problem is that when I test this code:
a=rand(2,1);
b=rand(2,1);
z=dotProd(a,b);
I get:
z=0
even though a and b are not orthogonal. I verified this with the MATLAB dot() function. I've picked over the code and can't quite seem to find where I'm going awry. Some suggestions would be appreciated.
Thank you.

Answer 1

That's because you're not returning the result of the dot product. z makes a local copy of itself in your dotProd function. Even though you are making modifications to z, those changes are not reflected because the scope of z inside dotProd is of local scope. You need to update your function that computes the dot product to return something. In addition, you are setting the output of the function before computing the dot product.
As such, do this:
// Change - Remove z as input
double dotProd(double *a, double *b, mwSize n)
{
mwSize i;
double z = 0.0; // Initialize z to 0.0
for(i=0;i<n;i++){
z+=a[i] * b[i];
}
return z; // Return z
}
Then simply do:
void mexFunction( int nlhs, mxArray *plhs[],
int nrhs, const mxArray *prhs[])
{
double z; //Output scalar - Change - don't need to initialize
double *b, *a; //Input vectors
int n;
a = mxGetPr(prhs[0]); //pointer to a
b = mxGetPr(prhs[1]); //pointer to b
n = mxGetM(prhs[0]);
// Create output
z = dotProd(a,b, (mwSize)n); // Change - returning output
plhs[0] = mxCreateDoubleScalar(z);
}
If you insist on changing z in the function and not letting the function return anything, you'll need to pass a pointer to z and change what z refers to. In other words, you would do this:
// Change - Make z point to a double
void dotProd(double *a, double *b, double *z, mwSize n)
{
mwSize i;
for(i=0;i<n;i++){
*z+=a[i] * b[i]; // Change - Refer to pointer
}
}
Now, do:
void mexFunction( int nlhs, mxArray *plhs[],
int nrhs, const mxArray *prhs[])
{
double z = 0.0;
double *b, *a; //Input vectors
int n;
a = mxGetPr(prhs[0]); //pointer to a
b = mxGetPr(prhs[1]); //pointer to b
n = mxGetM(prhs[0]);
// Create output
dotProd(a,b, &z, (mwSize)n); // Change - Pass pointer of z to function
plhs[0] = mxCreateDoubleScalar(z);
}
BTW, you still need to call dotProd before you set the output. That's why you kept getting 0 because z was 0 before you set the output, then you called dotProd after.