Hi,

I'm having problems with MatrixXd allocation (program crashes) when inside a parallel region, eg

Code: Select all

#pragma omp parallel
{
MatrixXd M(20,20);
... do some computations
}

I found out that if using a definied type like this

Code: Select all

typedef Matrix<double, Dynamic,Dynamic,RowMajor,20,20> Tensor;

instead MatrixXd, solve the problem. However, I have a lot of matrices
with very different sizes in my code, it would be necessary to create a type for each one, so this option is not feasible.

any suggestions?

Thank you

that's rather strange. which Eigen version do you have? we fixed an openmp issue in 3.0.5 and 3.1.0-alpha2 but I don't see how your trick could workaround this precise isssue. if updating does not solve your issue, could you provide a small exemple that trigger the issue... thanks.