CODEWITHSUNDEEP
c++matlabmex
Darkmoor
Darkmoor

Reputation: 902

mexfunction Inputs confusion

I am new in mex type programming, basically I am writting my first mexfunction, and I am encountering a silly problems. I attach a part of my code: 

void mexFunction(int nlhs, mxArray *plhs[],
    int nrhs, const mxArray *prhs[])
{

        int mrows,ncols;
        mrows = mxGetM(prhs[0]);
        ncols = mxGetN(prhs[0]);

        int numElements = ncols;
        size_t size = numElements * sizeof(float);

        float *in_A = (float *)mxMalloc(size);

        float *in_B = (float *)malloc(size);

        float *out_C = (float *)malloc(size);

        if (in_A == NULL || in_B == NULL ||out_C == NULL)
        {
            mexErrMsgTxt("Failed to allocate host vectors!\n");
        }

        if (nrhs != 2) 
            mexErrMsgTxt("Two inputs required. A vector and a multiplier");
        if (nlhs != 1) 
            mexErrMsgTxt("One output required. The resulting vector");


        in_A = (float *)mxGetPr(prhs[0]);
            in_B = (float *)mxGetPr(prhs[1]);

        //printf("%f %f\n", in_A[2], in_A[1]);
        //mexPrintf("%d\n", prhs[1]);

           plhs[0] = mxCreateDoubleMatrix(mrows,ncols, mxREAL);
           out_C = (float *)mxGetPr(plhs[0]);

            add_two_vectors(in_A,in_B,out_C);

}

The function add_two_vectors(float* A,float* B,float* C) is working but when I am running the script inside a mexfunction I have a problem. Also it reads correctly the nrows and ncols variables. The problem is that I can not insert the correct inputs in the function. I have tried to print some of the vector elements but it prints different elements than the input vector elements. I am calling mexfunction like that a = addVector([1.1 2.2 355],[2.5 45 5.5]). Can someone enlighten me??

Thank you in advance.

PS: Matlab 2011a, VS 2010, Win 7 64x.

Upvotes: 2

Views: 1581

Answers (2)

chappjc
chappjc

Reputation: 30589

You can't just static cast the output of mxGetPr (a double*) to a float* without misinterpreting the data. Read in the contents as double and convert each element, not the pointer.

Alternatively, you can input a float array with single(...) on the input in MATLAB, and use (float *)mxGetData instead of mxGetPr.

Upvotes: 0

Amro
Amro

Reputation: 124563

Let me give an example how I would implement such a function:

add_vectors.cpp

#include "mex.h"

void add_two_vectors(float *c, const float *a, const float *b, const size_t sz)
{
    for (size_t i=0; i<sz; i++) {
        c[i] = a[i] + b[i];
    }
}

void mexFunction(int nlhs, mxArray *plhs[], int nrhs, const mxArray *prhs[])
{
    // input validation
    if (nrhs != 2 || nlhs > 1) {
        mexErrMsgTxt("Wrong number of input/output arguments.");
    }
    if (!mxIsSingle(prhs[0]) || !mxIsSingle(prhs[1])) {
        mexErrMsgTxt("Inputs must be single arrays.");
    }
    if (mxIsComplex(prhs[0]) || mxIsComplex(prhs[1])) {
        mexErrMsgTxt("Inputs must be real arrays.");
    }
    if (mxIsSparse(prhs[0]) || mxIsSparse(prhs[1])) {
        mexErrMsgTxt("Inputs must be dense arrays.");
    }
    if (mxGetNumberOfElements(prhs[0]) != mxGetNumberOfElements(prhs[1])) {
        mexErrMsgTxt("Inputs must have the same size.");
    }

    // create ouput array
    mwSize numel = mxGetNumberOfElements(prhs[0]);
    mwSize ndims = mxGetNumberOfDimensions(prhs[0]);
    const mwSize *dims = mxGetDimensions(prhs[0]);
    plhs[0] = mxCreateNumericArray(ndims, dims, mxSINGLE_CLASS, mxREAL);

    // get pointers to data
    float *c = (float*) mxGetData(plhs[0]);
    float *a = (float*) mxGetData(prhs[0]);
    float *b = (float*) mxGetData(prhs[1]);

    // perform addition: c = a + b
    add_two_vectors(c, a, b, numel);
}

The function expects two arrays of type single of the same size (but can be of any dimensions including scalars, vectors, matrices, N-D arrays).

>> mex -largeArrayDims add_vectors.cpp
>> add_vectors(rand(4,'single'),rand(4,'single'))
ans =
    0.6204    1.2307    1.1657    0.7583
    1.3147    0.3817    0.3059    1.0377
    0.8142    0.3519    0.8619    1.1154
    1.0403    1.1578    0.1365    1.0048

Upvotes: 1

Related Questions