Hello,

I try to create threads with OpenMP:

ublas::matrix<T> l_matrix( rows, cols );

#pragma omp parallel for shared(l_matrix)

for(std::size_t n=0; n < l_matrix.size1(); ++n)

ublas::row(l_matrix, n) = <set a vector>

is this a problem with the ublas data structures something like

shared access to the elements or anything else?

Should I transform the the loop to:

#pragma omp parallel for shared(l_matrix)

for(std::size_t n=0; n < l_matrix.size1(); ++n)

#pragma omp critical

ublas::row(l_matrix, n) = <set a vector>

Thanks

Phil