Use parallel for_each for assembling data

Question

I wrote a finite element program which assembles small matrices into a huge sparse matrix. It is multithreaded (using std::thread). Depending on the number of t