Pardiso sparse solver has been successfully used to solve my problem through Eigen 3.3.3, but I cannot set the threads number used by the solver. For example, Pardiso solver only use four threads automatically for a computer with one processor four cores eight threads. So how to set the threads number used by Pardiso sparse solver manually.

What is the type of Eigen::EigenBase<Derived>::Index, these code will get the warming：data may be lost from Eigen::EigenBase<Derived>::Index through int.

Code: Select all

int eleNumber = 0;
int nodeNumber = 0;
Eigen::MatrixXd eleOutNode(100, 100);
eleNumber = eleOutNode.rows();
nodeNumber = eleOutNode.cols();

this warming will happen if the target is x64 architecture program with the visual studio 2017 IDE.

So sad...

It might be done by editing the proper entry of the array given by pardisoParameterArray(), then look at MKL's pardiso documention to find the right entry. But it could be that pardiso won't allow you to use your hyperthrea. This is for instance the case of dense matrix products routines that already use almost 100% of your CPU recourses so there is no room for additional hyperthreads.

Eigen::Index is defined as std::ptrdiff_t.

ggael wrote:It might be done by editing the proper entry of the array given by pardisoParameterArray(), then look at MKL's pardiso documention to find the right entry. But it could be that pardiso won't allow you to use your hyperthrea. This is for instance the case of dense matrix products routines that already use almost 100% of your CPU recourses so there is no room for additional hyperthreads.

Eigen::Index is defined as std::ptrdiff_t.

Get it, thanks.