Question

我正在尝试通过并行化加速Matlab程序中的某些处理。我无法访问Parallel Computing Toolbox，所以我一直试图通过mex文件来实现。我所拥有的（可能）相关部分是：

#include "mex.h"
#include "math.h"
#include <windows.h>
#include <process.h>

#define voidthread unsigned __stdcall
#define ThreadHANDLE HANDLE

...


voidthread computeloop (double **Args) {
   ...
}

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

   ...

   /* double pointer array to store all needed function variables) */
   double ***ThreadArgs, **ThreadArgs1;

   /* Handles to the worker threads */
   ThreadHANDLE *ThreadList;

   ...

   /* Reserve room for handles of threads in ThreadList  */
   ThreadList = (ThreadHANDLE*)malloc(Nthreads* sizeof( ThreadHANDLE ));
   ThreadID = (double **)malloc( Nthreads* sizeof(double *) );
   ThreadArgs = (double ***)malloc( Nthreads* sizeof(double **) );


   /* Assign pointer to output. */
   O_grad =(double *)mxGetData(plhs[0]);

   ... 

   for (i=0; i<Nthreads; i++) {

      /*  Make Thread ID  */
      ThreadID1= (double *)malloc( 1* sizeof(double) );
      ThreadID1[0]=(double)i;
      ThreadID[i]=ThreadID1;

      O_gradThread[i]=mxCreateNumericArray(3, Osize, mxDOUBLE_CLASS, mxREAL);

      /*  Make Thread Structure  */
      ThreadArgs1 = (double **)malloc( 10* sizeof( double * ) );
      ThreadArgs1[0]=O_grid_size;      
      ThreadArgs1[1]=Nthreadsd;
      ThreadArgs1[2]=d_p;
      ThreadArgs1[3]=dxa;
      ThreadArgs1[4]=dya;
      ThreadArgs1[5]=dza;
      ThreadArgs1[6]=d_p_size;
      ThreadArgs1[7]=O_grid_index;
      ThreadArgs1[8]=ThreadID[i];
      ThreadArgs1[9]=(double *)mxGetData(O_gradThread[i]);

      ThreadArgs[i]=ThreadArgs1;
      ThreadList[i]=(HANDLE)_beginthreadex(NULL, 0, &computeloop, ThreadArgs[i], 0, NULL);
   }

   ...
}

但是，当我尝试编译它时，我收到错误消息

error C2664: '_beginthreadex' : cannot convert parameter 3 from 'unsigned int (__cdecl *)(double **)' to 'unsigned int (__cdecl *)(void *)'

引用mexFunction中的行ThreadList[i]=(HANDLE)_beginthreadex(NULL, 0, &computeloop, ThreadArgs[i], 0, NULL);。

当我尝试更改computeloop以获取void *参数而不是double **（类似于this question的解决方案）并在其中投射参数时，我刚刚结束更多的错误。此外，这种通用模式适用于我在Matlab文件交换中看到的几个文件。这是否是其他地方出现错误的情况，直到此时编译器才会捕获，或者我错过了什么？我正在使用Microsoft Visual Studio 2010 Professional作为底层编译器通过Matlab编译它。

无法使用_beginthreadex编译MEX文件

0 个答案: