How would I reduce overhead in mex function?

Question

hotaka yagi 2018-1-3

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/375300-how-would-i-reduce-overhead-in-mex-function

评论： michio 2018-1-4

在 MATLAB Online 中打开

I want to know how to speed up mex function.

I checked the time of same c program of matrix-vector product, comparing to matlab-mex and without matlab.

the below is mex file program.

    #include "mex.h" 
    #include "math.h"
    /* The computational routine */ 
    void MatPro(double *x, double *y, double *z, mwSize m, mwSize n) 
    { 
        int i; 
        int j; 
        /* main program */ 
        for(i=0;i<m;i++){ 
            z[i]=0; 
            for(j=0;j<m;j++){ 
                z[i]=z[i]+x[m*j+i]*y[j]; 
                 } 
            } 
      } 
    /* The gateway function */ 
    void mexFunction( int nlhs, mxArray *plhs[], 
        int nrhs, const mxArray *prhs[]) 
    { 
        double *in1;              /* input scalar */ 
        mwSize n;                   /* size of matrix */ 
        mwSize m; 
        double *in2; 
        double *out;              /* output matrix */ 
        /* get the value of the input  */ 
        in1 =mxGetPr(prhs[0]); 
        /* create a pointer to the real data in the input matrix  */ 
        in2 =mxGetPr(prhs[1]); 
        /* get dimensions of the input matrix */ 
        m = (mwSize)mxGetM(prhs[1]); //GYO 
        n = (mwSize)mxGetN(prhs[1]); //RETU 
        /* create the output matrix */ 
        plhs[0] = mxCreateDoubleMatrix(m,n,mxREAL); 
        /* get a pointer to the real data in the output matrix */ 
        out = mxGetPr(plhs[0]); 
        /* call the computational routine */ 
        MatPro(in1,in2,out,m,n); 
    }

I experienced using the file in matlab. Also I checked the time.

    M=rand(32768); 
    x=rand(32768,1); 
    tic; 
    y=MatPro(M,x); 
    toc;

Second, c program of matrix-vector product is below

    #include "stdio.h" 
    #include "time.h" 
    #include "math.h" 
    #include "stdlib.h" 
    #include "omp.h" 
    void matVec(long *ans1, long *ans2) 
    { 
    // set variables 
    int m=32768; 
    int n=1; 
    int* x; 
    int* y; 
    int* z; 
    struct timespec startTime, endTime; 
    // set matrix 
    x=(int *)malloc(sizeof(int) * m*m); 
    y=(int *)malloc(sizeof(int) * m*n); 
    z=(int *)malloc(sizeof(int) * m*n); 
    // set value 
    for (i=0;i<m*n;i++) { 
        x[i]=rand(); 
        y[i]=rand(); 
    } 
    // check start time 
    clock_gettime(CLOCK_REALTIME, &startTime); 
    // calculation 
    for(i=0;i<m;i++){ 
        z[i]=0; 
        for(j=0;j<m;j++){ 
            z[i]=z[i]+x[m*i+j]*y[j]; 
        } 
    } 
    // check end time 
    clock_gettime(CLOCK_REALTIME, &endTime); 
    // output
    if (endTime.tv_nsec < startTime.tv_nsec) { 
        *ans1=endTime.tv_sec - startTime.tv_sec - 1; 
        *ans2=endTime.tv_nsec + (long int)1.0e+9 - startTime.tv_nsec; 
    } else { 
        *ans1=endTime.tv_sec - startTime.tv_sec; 
        *ans2=endTime.tv_nsec - startTime.tv_nsec; 
    } 
 } 
    int main(void){  
    long *ans1; 
    long *ans2; 
    printf("dimension = %09d, time = ",m);
    matVec(&ans1,&ans2); 
    printf("%5ld.%09ld",ans1,ans2); 
    printf("  (sec)\n"); 
    return 0; 
} 
I used the file without matlab to compare time.

I used Macbook Pro (intel core i7, 2.5GHz, 16GBmemory, macOS High sierra ver 10.13.1). The compiler is clang 5.0.0. I didn't use compiler option. Matlab is 2017R version.

The result is