Armadillo sparse real matrix multiplication with c

问题:

I'm trying to multiply a sparse real matrix with a complex vector but the program does not compile. If I change the vector to real or the matrix to dense, then everything goes through. A sample code is:

#define ARMA_64BIT_WORD
#include <armadillo>
#include <iostream>
#include <stdio.h>
#include <math.h>

using namespace arma;

int main(){

  size_t n(5);
  vec vR(randu<vec>(n)), vI(randu<vec>(n)); //Create random complex vector 'v'
  cx_vec v(vR, vI);

  std::cout<<"\n\tMultiplying real matrix with complex vector:"<<std::endl;
  mat R = randu<mat>(n,n);
  R*v; // -------------> COMPILES

 std::cout<<"\n\tMultiplying real sparse matrix with complex vector:"<<std::endl;
 sp_mat Rs = sprandu<sp_mat>(n,n,0.2);
 Rs*v; // ------------> DOES NOT COMPILE

 return 0;

}

Any recommendations for a solution? I'm using Armadillo version 5.200.1.

回答1:

I've had the same problem with multiplying two sparse matrices of different numeric type (see here). It seems that multiplying a sparse object with any other non-scalar type (be it sparse or dense) only works for equal numeric types for now. I am hoping they will implement that soon!

You can check by looking into operator_times.hpp at lines 502 and 454: the boolean expression in the enable_if2 template argument will evaluate to false if two objects with different template parameters are multiplied, hence eliminating that template from the list of candidates for the operator* overload.