MATLAB: Mex builds in Matlab, but crashes. Breakpoint not reached in Visual Studio debugging

arraysfor loopMATLABmex

Hello,
I've been trying to build a mex file to get some fast C++ computational routine running instead of matlab loops.
The mex builds OK, but matlab crashes when I try to use it in the M console. I just can't figure out how to debug it, because breakpoints are never reached wherever I put them in the cpp in Visual Studio. (Just to precise: I did the debugging with another version automatically provided by matlab coder, but it was slower than the original .m, that's why I tried to write it down by myself)
I'm a beginner with this kind of stuff, and tried to write the code following matlab help and examples given in matlab/extern. This code is about creating a 3D matrix output with computation that uses 3 * 3vectors (3 different sizes, corresponding to the 3 dimensions, and the 3 loops) as input.
Here it is, it is rather short. If someone can look at it and tell me if something looks wrong, any help would be greatly appreciated.
#include <cmath>
#include "mex.h"
/* The computational routine */
void comp_R(  double *Xobs, double *Yobs, double *Zobs,
        double *x, double *y, double *z,
        double *X, double *Y, double *Z, double *R,
        mwSize M, mwSize T, mwSize GS)
{
  mwSize m,t,gs;
  for(m=0; m < M ; m++)
  {
    for(t=0; t < T ; t++)
    {
      for(gs=0; gs < GS ; gs++)
      {
        R[gs*(T*M)+t*M+m]=sqrt((Xobs[gs]+(x[t]-x[0])-X[m])*(Xobs[gs]+(x[t]-x[0])-X[m])+(Yobs[gs]+(y[t]-y[0])-Y[m])*(Yobs[gs]+(y[t]-y[0])-Y[m])+(Zobs[gs]+(z[t]-z[0])-Z[m])*(Zobs[gs]+(z[t]-z[0])-Z[m])); 
        /*delay[gs][t][m]=R[gs][t][m]/c;*/
      }
    }
  }
}
void mexFunction(int nlhs, mxArray *plhs[], int nrhs, const mxArray *prhs[])
{
  /* check number of arg */
  if(nrhs!=9) {
  mexErrMsgIdAndTxt("MyToolbox:vec_geom:nrhs","Nine inputs required.");
  }
  if(nlhs!=1) {
  mexErrMsgIdAndTxt("MyToolbox:vec_geom:nlhs","One output required.");
  }
  /* variables initialisation*/
  
  double *xMatrix,*yMatrix,*zMatrix,*XMatrix,*YMatrix,*ZMatrix,*XobsMatrix,*YobsMatrix,*ZobsMatrix;  // input matrices
  double  *outMatrix; // output matrix  
  mwSize  n_GS,n_T,n_M;  // grid size, time samples, number of microphones  
  mwSignedIndex  dims[3];
  
  /* create a pointer to the real data in the input matrix and get dimensions of the input matrix  */
  XobsMatrix  = mxGetPr(prhs[0]); n_GS = mxGetN(prhs[0]);
  YobsMatrix  = mxGetPr(prhs[1]);
  ZobsMatrix  = mxGetPr(prhs[2]);
  xMatrix    = mxGetPr(prhs[3]);  n_T  = mxGetN(prhs[3]);
  yMatrix    = mxGetPr(prhs[4]);
  zMatrix    = mxGetPr(prhs[5]);
  XMatrix    = mxGetPr(prhs[6]);  n_M  = mxGetN(prhs[6]);
  YMatrix    = mxGetPr(prhs[7]);
  ZMatrix    = mxGetPr(prhs[8]);
  
  /* create output matrix */
  dims[0]    =   n_GS ;
  dims[1]    =   n_T ;
  dims[2]    =   n_M ;
  plhs[0]    =   mxCreateNumericArray(3, dims, mxSINGLE_CLASS, mxREAL);
  
  /* get a pointer to the real data in the output matrix */
  outMatrix =  mxGetPr(plhs[0]);
  
  /* call the computational routine */
  comp_R(XobsMatrix,YobsMatrix,ZobsMatrix,xMatrix,yMatrix,zMatrix,XMatrix,YMatrix,ZMatrix,outMatrix,n_GS,n_T,n_M);
}
I can give the error log if needed (stack trace that comes after "Abnormal termination: Segmentation violation" is just meaningless to me…)
Thanks, Raphaël

Best Answer

    double *outMatrix; // output matrix  
          :
    plhs[0] = mxCreateNumericArray(3, dims, mxSINGLE_CLASS, mxREAL);
          :
    outMatrix = mxGetPr(plhs[0]);
          :
comp_R(XobsMatrix,YobsMatrix,ZobsMatrix,xMatrix,yMatrix,zMatrix,XMatrix,YMatrix,ZMatrix,outMatrix,n_GS,n_T,n_M);
          :
void comp_R( double *Xobs, double *Yobs, double *Zobs,
      double *x, double *y, double *z,
      double *X, double *Y, double *Z, double *R,
      mwSize M, mwSize T, mwSize GS)
I haven't looked at your code in any real detail, but in the above extract you create plhs[0] as a single class variable (equivalent to a float), but the pointers you access the data with are double class. My guess is you are running off the end of the allocated memory with your double access in the comp_R routine. Assuming all of your loop indexing is correct (which I haven't looked at), maybe all you need to do is create plhs[0] as a mxDOUBLE_CLASS variable instead of a mxSINGLE_CLASS variable.
Related Question