在 MATLAB mexFunction 中,您如何将 GNU 科学图书馆矩阵和向量分配给 plhs?

In the MATLAB mexFunction how do you assign GNU Scientific Library matrices and vectors to plhs?

我有 MATLAB 代码,我想调用 GNU 科学库 (GSL) C 库来计算 singular value decomposition of a matrix. Basically, I need to implement the gateway MEX function。我知道如何设置输入变量,但如何设置输出变量?

到目前为止,这是我的代码:

build.m

mex -v -L/home/miran045/reine097/c-libs/lib/ -R2017b project_m.c -lgsl -lgslcblas -lm

call_c_from_matlab.m

A_data = [1.0, 0.0, 0.0, 0.0, 2.0, 0.0, 0.0, 3.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 2.0, 0.0, 0.0, 0.0];

[U, S, V] = project_m(4, 5, A_data);

project_m_from_c

#include <stdio.h>
#include <stdlib.h>
#include <math.h>

#include <gsl/gsl_math.h>
#include <gsl/gsl_matrix.h>
#include <gsl/gsl_vector.h>
#include <gsl/gsl_linalg.h>

#include "mex.h"
    
void run_svd(const size_t M, const size_t N, double A_data[], gsl_matrix * B, gsl_matrix * V, gsl_vector * S, gsl_vector * work)
{
  gsl_linalg_SV_decomp(B, V, S, work);
}

static double
get_double(const mxArray *array_ptr)
{
    double *pr;
    pr = mxGetPr(array_ptr);
    double result = (double) *pr;
    
    return result;
}

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

   /* You should put code here to check nrhs, nlhs, prhs[0], prhs[1], prhs[2] */

    int m = (int) get_double(prhs[0]);
    int n = (int) get_double(prhs[1]);
    double *mexArray =  mxGetPr(prhs[2]);
    int Bm, Bn; /* added */
   
  gsl_matrix * B;
  gsl_matrix * V;
  gsl_vector * S;
  gsl_vector * work;
  gsl_matrix_view A = gsl_matrix_view_array(mexArray, m, n);
  if (n > m) {
    Bm = n; Bn = m; /* added */
    work = gsl_vector_alloc(m);

    gsl_matrix_transpose_memcpy(B, &A.matrix);
  } else {
    Bm = m; Bn = n; /* added */
    work = gsl_vector_alloc(n);

    gsl_matrix_memcpy(B, &A.matrix);
  }

   plhs[0] = mxCreateDoubleMatrix( Bm, Bn, mxREAL );
   plhs[1] = mxCreateDoubleMatrix( Bn,  1, mxREAL );
   plhs[2] = mxCreateDoubleMatrix( Bn, Bn, mxREAL );
   
   run_svd(m, n, mexArray, plhs[0], plhs[2], plhs[1], work);
   
   gsl_vector_free(work);
}

但是我收到以下错误:

gsl: ../gsl/gsl_matrix_double.h:278: ERROR: first index out of range
Default GSL error handler invoked.
matlab&
[2] 130182
[1]   Killed                  matlab
reine097@native [/panfs/roc/groups/4/miran045/reine097] % MATLAB is selecting SOFTWARE OPENGL rendering.
MATLAB: smart_ptr.cpp:82: matrix::unique_mxarray_ptr matrix::from_matlab(mxArray* const&): Assertion `Input to from_matlab must have temporary or local scope' failed.
matlab&
[3] 115869
[2]   Segmentation fault      (core dumped) matlab
reine097@native [/panfs/roc/groups/4/miran045/reine097] % MATLAB is selecting SOFTWARE OPENGL rendering.
reine097@native [/panfs/roc/groups/4/miran045/reine097] % gsl: svd.c:79: ERROR: length of workspace must match second dimension of matrix A
Default GSL error handler invoked.
matlab&
[4] 89615
[3]   Killed                  matlab

您的代码存在多个问题:

  • 您没有检查输入是否完全符合预期(实数、双精度、非稀疏等)
  • 您正在将未使用 MATLAB API 函数分配的内存附加到 mxArray
  • 您在将指针附加到 mxArray 后立即释放指针后面的内存
  • 在使用 plhs[ ] mxArray 变量之前不要创建它们

通常,在使用 mex 例程时,人们会按顺序执行以下操作:

  • 检查输入是否完全符合预期
  • 创建输出 plhs[ ] mxArray 变量
  • 获取指向 plhs[ ] mxArray 变量的数据区域的指针
  • 将这些指针传递给计算例程

在您的情况下,您可能有理由希望使用 gsl_matrix_alloc( ) 例程而不是 MATLAB API 内存分配函数来进行内存分配。很好,但是如果你这样做,那么你将不得不 copy 结果到 plhs[ ] mxArray 数据区。您不能直接附加指针,因为这会搞砸 MATLAB 内存管理器并导致下游崩溃。但在所有情况下,您都需要先创建 plhs[ ] mxArray 变量。例如,如果我们只使用您当前的内存分配方案,您将需要复制数据:

编辑 gsl 数据复制

/* Create MATLAB mxArray from gsl_vector */
mxArray *gslvector2MATLAB(gsl_vector *V)
{
    mxArray *mx;
    double *data, *target;
    size_t n, stride;
    if( V ) {
        n = V->size;
        data = V->data;
        stride = V->stride;
        mx = mxCreateDoubleMatrix(n,1,mxREAL);
        target = (double *) mxGetData(mx);
        while( n-- ) {
            *target++ = *data;
            data += stride;
        }
    } else {
        mx = mxCreateDoubleMatrix(0,0,mxREAL);
    }
    return mx;
}

/* Create MATLAB mxArray from gsl_matrix */
mxArray *gslmatrix2MATLAB(gsl_matrix *M)
{
    mxArray *mx;
    double *data, *target;
    size_t i, j, m, n;
    if( M ) {
        m = M->size1;
        n = M->size2;
        data = M->data;
        mx = mxCreateDoubleMatrix(m,n,mxREAL);
        target = (double *) mxGetData(mx);
        for( i=0; i<m; i++ ) {
            for( j=0; j<n; j++ ) {
                target[i+j*m] = data[j]; /* tranpose */
            }
            data += M->tda;
        }
    } else {
        mx = mxCreateDoubleMatrix(0,0,mxREAL);
    }
    return mx;
}

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

   /* You should put code here to check nrhs, nlhs, prhs[0], prhs[1], prhs[2] */

    int m = (int) get_double(prhs[0]);
    int n = (int) get_double(prhs[1]);
    double *mexArray =  mxGetPr(prhs[2]);
   
  gsl_matrix * B;
  gsl_matrix * V;
  gsl_vector * S;
  gsl_vector * work;
  gsl_matrix_view A = gsl_matrix_view_array(mexArray, m, n);
  if (n > m) {
    B = gsl_matrix_alloc(n, m);
    V = gsl_matrix_alloc(m, m);
    S = gsl_vector_alloc(m);
    work = gsl_vector_alloc(m);

    gsl_matrix_transpose_memcpy(B, &A.matrix);
  } else {
    B = gsl_matrix_alloc(m, n);
    V = gsl_matrix_alloc(n, n);
    S = gsl_vector_alloc(n);
    work = gsl_vector_alloc(n);

    gsl_matrix_memcpy(B, &A.matrix);
  }

    run_svd(m, n, mexArray, B, V, S, work);
    
    nlhs = 3; // delete this line
    mxSetPr(plhs[0], B); // delete this line
    mxSetPr(plhs[1], S); // delete this line
    mxSetPr(plhs[2], V); // delete this line

    /* call custom routines to copy data as transpose */
    plhs[0] = gslmatrix2MATLAB(B);
    plhs[1] = gslvector2MATLAB(S);
    plhs[2] = gslmatrix2MATLAB(V);

  gsl_matrix_free(B);
  gsl_matrix_free(V);
  gsl_vector_free(S);
  gsl_vector_free(work);
}

我不熟悉 gsl_matrix 或 gsl_vector 类型,因此无法建议您如何复制上述数据。通常,您可以只从 gsl 类型获取数据指针,然后 memcpy 数据。如果您遵循我上面概述的步骤并首先创建 plhs[ ] 变量,然后将它们的数据指针直接传递给您的 svd 例程,则可以避免此数据复制并提高代码的性能。