matlab mex

最新推荐文章于 2021-05-02 08:20:00 发布

zzchust

最新推荐文章于 2021-05-02 08:20:00 发布

阅读量2.5k

点赞数 1

分类专栏： linux 文章标签： linux openmp matlab

本文链接：https://blog.csdn.net/zzchust/article/details/47175443

版权

linux 专栏收录该内容

3 篇文章 0 订阅

订阅专栏

linux matlab版本中OpenMP的使用

需要重点注意linux和window mex编译的参数不一样！
这里参考edgeBOX论文提供代码的参数选择
Next, please compile mex code from within Matlab (note: win64/linux64 binaries included):
mex private/edgesDetectMex.cpp -outdir private [OMPPARAMS] mex private/edgesNmsMex.cpp -outdir private [OMPPARAMS] mex private/spDetectMex.cpp -outdir private [OMPPARAMS] mex private/edgeBoxesMex.cpp -outdir private
Here [OMPPARAMS] are parameters for OpenMP and are OS and compiler dependent.
Windows: [OMPPARAMS] = '-DUSEOMP' 'OPTIMFLAGS="$OPTIMFLAGS' '/openmp"'
Linux V1: [OMPPARAMS] = '-DUSEOMP' CFLAGS="\$CFLAGS -fopenmp" LDFLAGS="\$LDFLAGS -fopenmp"
Linux V2:[OMPPARAMS] = '-DUSEOMP' CXXFLAGS="\$CXXFLAGS -fopenmp" LDFLAGS="\$LDFLAGS -fopenmp"
To compile without OpenMP simply omit [OMPPARAMS]; note that code will be single threaded in this case.

对于windows matlab2014b版本，同样可以使用一下参数进行编译
mex euc_nn_mex_K.cpp COMPFLAGS="/openmp $COMPFLAGS"
但是，在linux上不能使用该参数，否则即使编译通过，线程数也只为1。
在不明确定义线程数的情况下，系统运行可以达到最大线程数，即系统核心数 CPU数目单个CPU核心数*。

另外还可以参考

mex euc_nn_mex.cc CXXFLAGS="\$CXXFLAGS -fopenmp -Wall" LDFLAGS="\$LDFLAGS -fopenmp";
mex kmeans_iter_mex.cc CXXFLAGS="\$CXXFLAGS -Wall"

代码参考

#include <assert.h>
#include <math.h>
#include "mex.h"
#include <sys/time.h>

void mexFunction(int nlhs, mxArray* plhs[],
                 int nrhs, const mxArray* prhs[]) {
  if (nrhs != 2) 
    mexErrMsgTxt("There should be two input arguments.");

  if (nlhs != 2 && nlhs != 1)
    mexErrMsgTxt("There should be one or two output arguments.");

  int p = mxGetM(prhs[0]);
  int n = mxGetN(prhs[0]);
  int nq = mxGetN(prhs[1]);

  if (mxGetM(prhs[1]) != (unsigned int)p)
    mexErrMsgTxt("Dimensionality of the two input matrices are not "
                 "consistent.");

  if (mxGetClassID(prhs[0]) != mxSINGLE_CLASS 
      || mxGetClassID(prhs[1]) != mxSINGLE_CLASS )
    mexErrMsgTxt ("Only single precision matrices are supported."); 

  float * b = (float*) mxGetPr(prhs[0]);  /* database vectors */
  float * v = (float*) mxGetPr(prhs[1]);  /* query vectors */

  plhs[0] = mxCreateNumericMatrix(nq, 1, mxINT32_CLASS, mxREAL);
  int * ind = (int*) mxGetPr(plhs[0]);

  float * dis = NULL;
  if (nlhs >= 1) {
    plhs[1] = mxCreateNumericMatrix(nq, 1, mxSINGLE_CLASS, mxREAL);
    dis = (float*) mxGetPr(plhs[1]);
  }

  int j = 0;
#pragma omp parallel shared(j)
  {
#pragma omp for
    for (j = 0; j < nq; j++) {
      float * vj = v + j * p;  // vj points to the j-th query point.
      float min_dis = 1e10;
      int best_ind = -1;
      float * bi = b;  // bi points to the i-th data point.
      // The pointer changes with the for variable i.
      for (int i = 0; i < n; i++, bi += p) {
        float disi = 0;
        for (int k = 0; k < p; k++) {
          disi += (bi[k] - vj[k]) * (bi[k] - vj[k]);
        }
        if (disi < min_dis) {
          min_dis = disi;
          best_ind = i;
        }
      }
      if (nlhs >= 1)
        dis[j] = min_dis;
      ind[j] = best_ind + 1;  // one-based
      if (best_ind == -1)
        mexErrMsgTxt ("Something is wrong, ind is still -1.");
    }
  }
}