Hello,
I don't understand the use of the function .eval() in the case of both dense and sparse matrix.
Is it use to avoid aliasing?
When must I use it and when is it not required?

Thanks

eval() permits to explicitly introduce a temporary in an expression. It's extremely rare to have to use it. With dense matrices, you can use to deal with an aliasing issues. Sometime it's also needed to send the result of an expression to a function expecting a raw pointer to the data, e.g. with OpenGL:

glVertex4fv((a+b).eval().data());

it is still not so clear to me.
I wrote this functions for summing sparse matrices:

Code: Select all

void Analysis::sumMatr(std::vector<SubDomain *> const &pSubD, SparseMatrix<complex<double> > &result){
   SparseMatrix<complex<double> > KMDsum(dimSys,dimSys);
   vector <SubDomain *>::const_iterator iter=pSubD.begin();;
   while (iter != pSubD.end()){
      KMDsum = (KMDsum + *((*iter++)->getKdyn())).eval();
   }
   result=KMDsum;
}

and

Code: Select all

void Analysis::sumMatr(SparseMatrix<complex<double> K, SparseMatrix<complex<double> M, SparseMatrix<complex<double> > &result){
   result=K+M;
}

in the first case I use .eval() while it is not required in the secondo case.

Are they both correct?

even in the first case, it should not be needed. What's the type returned by getKdyn() ?

ggael wrote:even in the first case, it should not be needed. What's the type returned by getKdyn() ?

it is SparseMatrix<complex<double> >.
It is simply a summing of sparse complex matrices